Identifying association anomalies of critical anomalies in network
By constructing network graphs and applying graph queries through an intent-based network management system, key anomalies and their related anomalies in the network can be quickly identified and resolved. This solves the problem of low efficiency in identifying and handling network anomalies in existing technologies, and improves the efficiency of network management and problem-solving capabilities.
Patent Information
- Application Number
- CN202511426089.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2025-09-17
- Filing Date
- 2025-09-30
- Publication Date
- 2026-03-31
AI Technical Summary
Existing network management systems struggle to quickly identify critical anomalies and their related anomalies when recognizing and handling network anomalies. This forces operators to spend a significant amount of time reviewing and investigating numerous alerts, leading to alert fatigue and inefficient problem-solving.
An intent-based network management system is adopted to identify and map key anomalies and their associated anomalies by constructing network graphs and applying predefined graph queries. It provides rapid anomaly analysis and solutions by leveraging graph query language and knowledge cards.
It enables rapid identification and resolution of critical anomalies and their related anomalies in the network, improves the efficiency of network management, reduces alarm fatigue, ensures network operation complies with Service Level Agreements (SLA/SLE), and provides a comprehensive view of network problems.
Smart Images

Figure CN121771053A_ABST
Abstract
Description
[0001] Related applications
[0002] This application claims the benefit of U.S. Patent Application No. 19 / 331,637, filed September 17, 2025, and U.S. Provisional Patent Application No. 63 / 701,474, filed September 30, 2024. Technical Field
[0003] This disclosure relates to computer networks, and more specifically, to anomalies in network systems. Background Technology
[0004] A computer network is a collection of interconnected computing devices that can exchange data and share resources. Various devices operate to facilitate communication between these computing devices. For example, a computer network may include routers, switches, gateways, firewalls, and various other devices to provide and facilitate network communication. In some cases, a computer network can be implemented in a data center with hundreds or even thousands of network devices as part of the network.
[0005] A Network Management System (NMS) enables administrators to monitor, configure, and manage network devices. The interaction between the NMS and the network configures network devices, ensuring the network is configured as needed, operates correctly, and is maintained efficiently. After discovering devices and establishing communication, the NMS can perform network configuration tasks. These tasks are executed based on the network administrator's policies, rules, or specific commands. Configuration tasks may include configuring devices involving application profiles or templates for routers, switches, firewalls, etc., and may include setting Internet Protocol (IP) addresses, Virtual Local Area Networks (VLANs), Access Control Lists (ACLs), routing protocols, or settings for their dedicated devices. Configuration tasks may also include configuring network policies such as Quality of Service (QoS), traffic prioritization, security rules, and firewall policies. Configuration tasks may also include setting services such as Dynamic Host Configuration Protocol (DHCP), Domain Name Service (DNS), Network Time Protocol (NTP), and load balancers.
[0006] NMS can also participate in monitoring and telemetry collection, thereby monitoring the network status after configuration to ensure that devices remain healthy and operate as expected. For example, telemetry data may include information about device health (e.g., CPU utilization, memory utilization, temperature), network traffic statistics (e.g., bandwidth utilization, packet drop, error rate), and link status (e.g., interface enable / failure status, port errors). As part of network monitoring, NMS can also perform configuration verification to ensure that the actual configuration state of the network matches its expected configuration state. If NMS detects any discrepancies or configuration problems, it can take action to align the actual network configuration state with the expected configuration state.
[0007] NMS can also interact with the network by monitoring events and generating alerts based on predefined thresholds or conditions. For example, NMS can trigger an alert to the network administrator if a link is down, traffic exceeds a limit, or a device approaches its resource capacity. Such events may be referred to as “anomalies” in this document. In some cases, NMS can automatically perform predefined actions when certain alerts are triggered, such as rerouting traffic or adjusting QoS settings. Summary of the Invention
[0008] In general, this disclosure describes techniques for analyzing anomalies in a network. In one aspect of this disclosure, the techniques relate to identifying anomalies associated with key anomalies in a network. A network management system can implement intent-based networking (IBN) to manage the network using a network graph that models the network's configuration and operational state.
[0009] In some aspects of these technologies, the network management system identifies multiple anomalies in the network that deviate from the network's intent. Each anomaly is associated with one or more nodes in the network graph, for example, as an attribute or "label" of the node. These can be stored in a 'key:value' format. The network graph can model the network intent. An analysis system, which could be the network management system or another computing system, applies predefined graph queries to the network graph to identify nodes with anomalies, the relationships between those nodes, and to match the anomalies. The predefined graph queries are associated with data indicating which of the anomalies matched by the predefined graph queries is the critical anomaly. This data can further indicate other anomalies associated with the critical anomaly. A critical anomaly is, for example, an anomaly that causes other anomalies associated with the critical anomaly, an anomaly that has the greatest impact among the anomalies matched by the predefined graph queries, or an anomaly that is otherwise considered important (i.e., "critical") by the operator or expert. The analysis system can execute predefined graph queries against the network graph and output an indication that the network is experiencing a critical anomaly when a match is found with a node in the network graph. The match can be a subgraph of the network graph. The predefined graph queries and data can be generated, stored, and displayed as knowledge cards.
[0010] The aforementioned technologies can provide one or more technical advantages with one or more practical applications. For example, identifying critical and associated anomalies from multiple anomalies allows operators to quickly identify high-value areas for investigation. Instead of reviewing and investigating hundreds or even thousands of anomalies, leading to alert fatigue, the analytics system presents operators with one or more critical anomalies, and once these critical anomalies are investigated and fixed, those associated with them may also be fixed. This allows operators or another system to resolve network-related issues more quickly. The technology provides a comprehensive view of problems and impacts on applications / services running on the network and facilitates the differentiation of which anomalies are side effects of critical anomalies or which are unrelated to critical anomalies. In other words, the technology using knowledge cards described herein can help improve the technical field of network management. For example, this technology can help allow users of network management systems to resolve problems within the network more quickly. This can include reconfiguring the network to address such problems.
[0011] In some technical aspects, the analysis system maps critical anomalies present in the network to problems at the application level. Services executing on compute nodes and clients interacting with these services via network connections may be affected by critical anomalies. The analysis system stores service impact data indicating one or more services that may be affected by critical anomalies. For example, a down interface can prevent access to services running on compute nodes that communicate using that interface. When identifying critical anomalies, the analysis system uses service impact data to identify one or more services that may be affected. The analysis system can output an indication of these one or more services. In some examples, the analysis system extends a visual depiction of at least a portion of the network topology to include service topology, effectively expanding the network graph to visually indicate the services and / or clients affected by critical anomalies. Returning to the example above, the service impact data associated with the critical anomaly indicates that a down interface of a leaf switch may affect all services running on compute nodes connected to the down interface of the leaf switch. The analysis system extends the network topology to indicate the services running on the affected compute nodes, and in some cases, the clients connected to those services. These indications of affected services and clients can be considered synthetic anomalies because they are not identified by the network management system using telemetry, configuration, or operational data from the network or compute nodes, but rather based on service impact data that is identified as potentially caused by critical anomalies.
[0012] The aforementioned technology can provide one or more technical advantages with one or more practical applications. For example, this technology can enable operators or systems to quickly identify affected services and take action to remediate them. This may include re-providing the affected service to another compute node, prioritizing the resolution of critical anomalies based on the priority of the affected services, or other actions. Resolving critical anomalies may include reconfiguring the network. Affected services are identified based on data obtained from the network, without relying on fault indications provided by the service itself. In some cases, this can provide earlier problem indications and clearly identify the problem as occurring within the network, rather than by the service itself or the compute node executing the service.
[0013] In some technical aspects, the network management system associates operational data with one or more nodes in the network graph. For example, such operational data might indicate faulty interfaces, suspended / cold interfaces, interface oscillations, poor fiber optic connections, latency issues, resource utilization, environmental factors (fans, power, temperature), device traffic, configuration discrepancies, the number of routes in an Ethernet Virtual Private Network (EVPN), the size of the EVPN flood list, etc. As an example, the CPU utilization for a leaf switch might be 80%, the link might have a latency >1ms, or the EVPN flood list might contain 25 interfaces. Operational data is associated with one or more nodes in the network graph, for example, as node attributes or "labels." An analysis system (which could be the network management system or another computing system) applies predefined graph queries to the network graph, matching nodes with attributes that satisfy thresholds defined in the graph query, the relationships between those nodes, and the attributes that satisfy the thresholds. Predefined graph queries might be based on the network's Service Level Agreement (SLA) / Service Level Expectations (SLE). These predefined graph queries are associated with data that the analysis system uses to identify and indicate one or more affected nodes in the network. The analysis system can perform predefined graph queries on the network graph and, when a match is found with a node in the network graph, output an indication that one or more affected nodes in the network are experiencing a problem (e.g., poor / inadequate health), or a positive indication that one or more "affected nodes" meet SLA / SLE requirements. This technology can be used to identify issues related to service health, link health, system / device health, EVPN network architecture health, or to demonstrate satisfactory network operation. Matches can be subgraphs of the network graph. Predefined graph queries and data can be generated, stored, and displayed as knowledge cards.
[0014] The technologies described above can provide one or more technical advantages with one or more practical applications. For example, these technologies can enable operators to quickly identify health problems in the network. This allows operators or another system to resolve network-related issues more quickly and ensures compatibility with SLAs / SLEs.
[0015] In the example, a system includes: a memory storing graph queries for a network and a network graph, wherein the network graph includes one or more nodes having one or more attributes indicating multiple anomalies of the network; and one or more processors coupled to the memory, wherein the memory stores instructions that, when executed, cause the one or more processors to: perform a graph query on the network graph of the network to determine a matching subgraph of the network graph, wherein the graph query matches one or more nodes and one or more attributes; and based on the determination of the matching subgraph, output an indication of associations between multiple anomalies.
[0016] In the example, one method includes the system acquiring a graph query and a network graph for a network, wherein the network graph includes one or more nodes having one or more attributes indicating multiple anomalies of the network; the system performing a graph query on the network graph to determine a matching subgraph of the network graph, wherein the graph query matches one or more nodes and one or more attributes; and based on the determination of the matching subgraph, outputting an indication of the association between multiple anomalies.
[0017] In the example, the non-transitory computer-readable storage medium includes instructions that, when executed, cause processing circuitry to: acquire a graph query and a network graph for a network, wherein the network graph includes one or more nodes having one or more attributes indicating multiple anomalies of the network; perform a graph query on the network graph of the network to determine a matching subgraph of the network graph, wherein the graph query matches one or more nodes and one or more attributes; and based on the determination of the matching subgraph, output an indication of associations between multiple anomalies.
[0018] Details of one or more examples are set forth in the accompanying drawings and the following description. Other features, objects, and advantages will be apparent from the specification, the drawings, and the claims. Attached Figure Description
[0019] Figure 1 This is a block diagram illustrating an example of a network managed using a network management system and an analysis system according to the technology disclosed herein.
[0020] Figure 2 This illustrates the technology used according to the present disclosure for... Figure 1 A block diagram of the analysis system and exemplary component set of the network management system.
[0021] Figure 3A and Figure 3B This is a conceptual diagram illustrating an exemplary network device communicating with a network management system according to the technology of this disclosure.
[0022] Figure 4 This illustrates the technology used according to the present disclosure for... Figure 1 A conceptual diagram of a network.
[0023] Figure 5 An exemplary user interface depicting a network diagram of a network according to the technology of this disclosure is shown.
[0024] Figures 6A to 6D A user interface generated according to the technology disclosed herein is depicted.
[0025] Figures 7A to 7C A network graph and a graph query subgraph according to the technology of this disclosure are depicted.
[0026] Figure 8 This is an exemplary diagram query based on the technology disclosed herein.
[0027] Figure 9 The technology according to this disclosure is described and Figure 8 The graph query provides an example subgraph of the network graph that matches the given graph.
[0028] Figure 10 The properties of nodes that have been augmented with abnormal data according to the techniques disclosed herein are listed.
[0029] Figure 11 A user interface generated by an analysis system according to the technology disclosed herein is depicted.
[0030] Figure 12A A user interface generated by an analysis system according to the technology disclosed herein is depicted.
[0031] Figure 12B The properties of nodes that have been augmented with abnormal data according to the techniques disclosed herein are listed.
[0032] Figure 13A A user interface generated by an analysis system according to the technology disclosed herein is depicted.
[0033] Figure 13B The properties of nodes that have been augmented with abnormal data according to the techniques disclosed herein are listed.
[0034] Figure 14 This is an exemplary system that implements the analysis system and network management system in more detail according to the technology disclosed herein.
[0035] Figure 15 This is an exemplary system that implements the analysis system and network management system in more detail according to the technology disclosed herein.
[0036] Figure 16 This is a flowchart illustrating an exemplary operating mode of an analysis system based on the technology disclosed herein.
[0037] Throughout the text and accompanying figures, the same reference characters refer to the same elements. Detailed Implementation
[0038] Intent-based networking is a software-enabled, automated process that uses a high level of intelligence, analytics, and coordination to improve network operation and uptime. When operators describe the business outcomes they wish to achieve, the network management system translates these goals into the configurations needed to achieve them, without requiring manual coding and execution of individual tasks.
[0039] For example, consider the need for secure communication between two networks. An intent would broadly declare the need for a secure tunnel between network A and network B. The operator would identify which traffic should use that tunnel and describe any other desired general properties of the tunnel. However, the operator does not need to specify how the tunnel is implemented, such as the number of devices to be used, how Border Gateway Protocol (BGP) advertising should be performed, or which specific features and parameters should be enabled. Instead, an intent-based network system can automatically generate a complete configuration for all devices based on the service description. The intent-based network system can then use closed-loop verification to continuously check the consistency between the network's intended state and its operational state, thereby continuously verifying the correctness of the configuration.
[0040] Intent-based networking (IBN) is a declarative network operation model. It contrasts with traditional imperative networking, which requires network engineers to specify the required sequence of operations on individual network elements, making it prone to errors. Traditionally, network operation has been driven by manual, command-line interface (CLI)-based operations, basic element management systems (EMS), or automated scripts. Most network outages are caused by human error during these network operations. Intent-based networking (IBN) reduces errors and risks while improving operational efficiency in several ways. For example, IBN validates intent objects before applying them to the network. An intent object is a high-level representation of the desired property or result achieved using the network. Validation is syntactic and includes semantic checks against the network-wide policy. IBN facilitates rapid rollback or rollforward. If a problem occurs during deployment push, the operator simply applies the appropriate version of the intent object to return to a known good state. IBN limits the impact and scope of failures during new intent exits through well-defined policies. IBN enables intent-based rollback. When a system knows the expected results for a particular configuration, it can achieve the same results by reconfiguring other network elements or using different mechanisms, and even maintain these results in the face of interruptions or device errors.
[0041] Network coordination systems can utilize intent-based networking for potentially mission-critical and large-scale deployments. Intent-based networking can significantly reduce the time to deliver reliable services from days or weeks to minutes, and help address operational challenges after infrastructure deployment. Intent-based networking can also involve intent guarantees. Through intent-based analytics, the network can maintain its initial intent throughout the service lifecycle. Figure 1Intent-based analytics can provide insights into network services, enabling teams to view the network as a complete service. Using analytics, intent-based networking enables faster root-cause analysis (RCA) and identification.
[0042] Figure 1 This is a block diagram illustrating an example of a network 2 managed by a network management system 10 and an analysis system 17 according to the technology of this disclosure. The network management system 10 described herein implements an intent-based network and can also perform intent-based analysis.
[0043] Network devices 14A to 14G (collectively referred to as "Network Device 14") in Network 2 are interconnected via communication links to form a communication topology for exchanging resources and information. For example, Network Device 14 may include routers, switches, gateways, bridges, hubs, access points, servers, firewalls or other intrusion detection systems (IDS) or intrusion prevention systems (IDP), computing devices / hosts / servers / nodes, computing terminals, printers, other network devices, or combinations of such devices. Although described herein as transmitting, conveying, or otherwise supporting data packets, network devices within Network 2 may transmit data according to any other discrete data unit defined by any other protocol. The communication links that interconnect Network Device 14 may be physical links (e.g., fiber optic, copper, etc.), wireless, or any combination thereof.
[0044] Network 2 can represent a data center network that connects physical infrastructure to network device 14. Generally, a data center network is a structured system of networked devices, protocols, and infrastructure designed to support the computing, storage, and communication needs of a data center. Data center hosting provides computing and storage systems for applications, data processing, and services to enterprises, cloud providers, and internet-based services. Figure 1 In the example, the physical infrastructure includes servers 11A through 11N (collectively referred to as "Server 11"). Server 11 may include compute servers that host applications and services deployed using, for example, virtual machines, containers, or other virtual computing instances or workloads. Server 11 may also include storage servers with one or more storage systems. Server 11 is connected to network 2 via a physical interface of a network interface card (NIC), and network 2 interconnects the compute servers and storage servers of Server 11 to enable data communication between Server 11 and distributed applications and storage.
[0045] In a data center network, network device 14 can be configured as a data center network architecture to interconnect servers 11 in one or more data centers. The switches of network device 14 may include top-of-rack switches, leaf switches, and backbone switches. A multi-tiered architecture can be used to build a data center network to manage large volumes of internal (east-west) and external (north-south) traffic. For example, a multi-tiered architecture can be a leaf-backbone architecture or a three-tier design.
[0046] Server 11 executes applications to provide services. Example services may include infrastructure services such as Domain Name Service (DNS), Dynamic Host Configuration Protocol (DHCP), authentication and directory services, backup and storage management, and load balancing. Other example services may include external services or client-facing services provided to tenants or clients; such services may include, for example, enterprise applications, web / email hosting, cloud computing services (e.g., compute, storage, containers, application hosting), virtualization services (e.g., virtual machine [VM] hosting), application servers, streaming services, collaboration and communication platforms, DevOps, backup and disaster recovery, content delivery networks, and e-commerce and other financial services.
[0047] Server 11 and / or Network 2 can implement network virtualization to abstract the physical networking infrastructure and create virtual network environments. Network virtualization allows for better resource allocation, scalability, and automation. For example, network device 14 and / or server 11 can be configured to implement a virtual network overlay that supports features such as virtual switches, virtual firewalls, and virtual routers to interconnect virtual computing instances or other workloads running on server 11. Virtualization reduces reliance on physical hardware, thus allowing for greater flexibility in managing workloads and traffic. Unless described in the context, Network 2 should be considered to include server 11.
[0048] Network 2 is shown connected to network 18 via one or more communication links. Network 18 can provide access to resources of server 11 by other devices. Network 18 can be a public network, such as the Internet, a private network, a VPN, or other networks. Network devices 14 can communicate with each other, server 11, and network 18 using various protocols (such as Border Gateway Protocol (BGP) or other routing protocols, Virtual Extensible LAN (VXLAN), Ethernet VPN or BGP-EVPN, Layer 2 protocols, etc.) at different layers of the Open Systems Interconnect model.
[0049] The network management system 10 is communicatively connected to the network device 14 via network 2. The network management system 10 can be directly or indirectly connected to different network devices 14. Once the network device 14 is deployed and activated, the administrator 12 uses the network management system 10, for example, using a device management protocol, to manage and monitor the network device. The administrator 12 can be a human operator or a computing system.
[0050] The network management system 10 (also referred to herein as the network management system (NMS)) and network devices 14 can be centrally maintained by a management group (such as an enterprise or vendor's IT group). An administrator 12 interacts with the network management system 10 to remotely configure, monitor, and analyze network devices 14. For example, the administrator 12 can receive alerts from the network management system 10 regarding any network device 14. Alerts may include alerts regarding abnormal operation of one or more network devices among those detected using the techniques described herein. The administrator 12 can also view the configuration data of network device 14, modify the configuration data of network device 14, add new network devices to network 2, remove existing network devices from network 2, or otherwise manipulate network 2 and the network devices therein. While networks have been described, the techniques disclosed herein can be applied to other public and private network types, including LANs, VLANs, VPNs, etc.
[0051] Administrator 12 can use network management system 10 to configure network device 14 to specify operational characteristics that further advance administrator 12's objectives. For example, administrator 12 can specify specific operational policies for network device 14 regarding security, device accessibility, traffic engineering, Quality of Service (QoS), Network Address Translation (NAT), packet filtering, packet forwarding, rate limiting, or other policies. Network management system 10 uses one or more network management and automation protocols designed to set configuration data in network device 14 and acquire telemetry data indicating the operational status of the network device. Such protocols may include Simple Network Management Protocol (SNMP), Network Configuration Protocol (NETCONF) or RESTCONF, OpenFlow / P4 or other protocols used in software-defined networking (SDN), telemetry protocols such as gRPC, etc. Network management system 10 may employ one or more automation frameworks that interact with network device 14 via Secure Shell (SSH) or Representational State Transfer (REST) APIs to automate the deployment and configuration of network 14. Network management system 10 and network device can communicate using communication 15 according to the aforementioned protocols.
[0052] User configurations for a device can be referred to as "intents." Intent-based network systems help administrators describe the expected network / compute / storage state. In some respects, user intents can be categorized as business policies or stateless intents. Business policies or stateful intents can be resolved based on the current state of the network. Stateless intents can be a completely declarative way of describing the expected network / compute / storage state, regardless of the current network state.
[0053] In some respects, stateful intents can include intents regarding anomaly detection within the network. Such intents can be referred to as anomaly detection intents. As an example, if the operational characteristics of a network device change from a baseline value established as described herein by exceeding a user-specified threshold, an administrator can express an intent to report an anomaly regarding the network device. This intent can be applied to a single network device or a group of network devices. Examples of such groups include network devices of the same brand and model, network devices from the same vendor, network devices in the same region, etc.
[0054] The network management system 10 enables intent-based networking to automate the management of network 2 in an intent-based manner. An administrator 12 defines how to configure and operate the network (intent 7), and the network management system 10 ensures that the network configuration and operation match intent 7. The network management system 10 models the representation of network 2 as a network graph 13, in which network devices 2, links, interfaces, and other network components are nodes, and the relationships or connections between nodes are edges. Edges can therefore represent, for example, physical cabling, logical links, protocols, or data flows. Network graph 13 is a graph-based data model that allows users to visualize and manage the entire network holistically. Figure 4 Network graph 700 illustrates an example of a network graph and is described in further detail below. A graph database (graphDB) can be used to store network graphs 13, which can be queried using a graph query language.
[0055] Using network diagram 13 to model network 2, network management system 10 enables a visual representation of the state of network 2, providing insights into how devices and services are connected. This structure allows for a comprehensive view of the network as a whole, visualizing relationships between devices, data flow paths, and dependencies between different components of network 2.
[0056] Administrator 12 uses network management system 10 to specify high-level intent 7 for network 2. Intent 7 for network 2 is high-level configuration data that describes and / or defines the expected results of network 2's architecture, configuration, and operation, rather than specific configuration details. For example, instead of configuring individual network devices, administrator 12 may specify isolation for specific workloads or that certain traffic should be load-balanced. Intent 7 for network 2 can be specified by administrator 12 using network management system 10 in various ways. For example, intent 7 can be expressed as structured input parameters, for example, according to YANG, JavaScript Object Notation (JSON), or other data modeling languages. Network management system 10 provides application programming interfaces (APIs), CLIs, or other methods through which administrator 12 can specify, interact with (e.g., query), and update intents.
[0057] In some examples, Intent 7 is specified as a template or model (also known as a "blueprint"). This intent may include the physical topology of the layout of network device 14, server 11, and the links between these devices; the logical topology defining how the network is logically segmented (e.g., subnets, VLANs, and routing policies) and how traffic is logically routed between network device 14 and server 11; intent-based policies specifying requirements such as security, performance, or compliance; and / or the roles of network device 14 or other network components (e.g., "backbone switch," "leaf switch," "link"), and the relationships between network device 14 or other network components. Intent 7 can be a network graph ("intent network graph"). Intent 7 can be a directed acyclic graph. Intent 7 can be queried using a graph query language.
[0058] Network management system 10 can use intent 7 to generate a corresponding network diagram 13 to represent the implementation of intent 7. Although intent 7 is a high-level specification, network diagram 13 captures operational details of network 2, such as device configuration, link status, and data flow. Network management system 10 converts the high-level specification into low-level configuration data, for example, for network device 14, and uses this low-level configuration data in a manner thus based on intent 7 to configure network 2, ensuring that the actual network topology and configuration are consistent with what is specified. That is, using network diagram 13, network management system 10 ensures that the actual state of network 2 is consistent with the expected state specified by intent 7. Network management system 10 checks and verifies that all devices are configured and operated according to the defined intent. Changes in the network determined based on configuration or telemetry data obtained from network 2 are reflected in network diagram 13 in real time, and network management system 10 can automatically respond to deviations by making corrections to bring network 2 into alignment with intent 7 or by notifying administrator 12.
[0059] The network management system 10 using network diagram 13 can perform closed-loop automation, where network 2 is continuously monitored and adjusted to meet expected conditions without human intervention. Through network diagram 13, the network management system 10 can continuously verify network performance, reduce misconfigurations, and ensure compatibility with design strategies.
[0060] Network graph 13 can be queried by administrator 12, for example, using network management system 10 or another system. Network graph 13 is continuously updated to reflect the real-time state of the network, allowing administrator 12 to perform graph queries to gain deeper insights into the status and relationships of network devices 14. Graph queries are based on relationships between nodes, such as finding paths between two devices or determining how services flow through the network. Graph queries can thus help administrator 12 perform various tasks, such as troubleshooting, monitoring, and configuration changes. For example, graph queries can achieve topology discovery by traversing network graph 13 to retrieve the entire network topology, including all devices and their interconnections. This provides visibility into how all switches, routers, and links are connected. Graph queries can be used to find all devices and links between server 11A and network device 14C. Graph queries can retrieve information indicating the status of all or a subset of network devices and links in network graph 13. Other graph queries may include those related to bandwidth and resource utilization, redundancy and resilience, or policy compliance. Graph queries can be expressed using GraphQL, Cypher, Gremlin, SPARQL, Attribute Graph Query Language (PGQL), or other supported languages to extract specific data or insights from the network graph (13). Graph queries can be executed internally via a REST API or via another type of interface.
[0061] The network management system 10 identifies anomalies in network 2. Typically, an anomaly is a deviation in network 2 from intent 7 (i.e., the expected network configuration or operating state). For example, the network management system 10 may identify anomalies based on configuration data or telemetry data obtained from devices in network 2, or based on probe data generated from probe network device 14 or server 11.
[0062] Telemetry data can be operational temperature data, voltage data, current draw data, or other operational characteristics relating to the operation of network device 14. Other features that can be collected include transmitted / received bytes / data packets indicating traffic, error packet counts such as Cyclic Redundancy Check (CRC), Frame Check Sequence (FCS), etc., which can indicate deteriorating operational status. Network management system 10 can analyze and use telemetry data in various ways. During the initial baseline establishment period, network management system 10 can collect and store telemetry data. In some aspects, the baseline establishment period can be thirty days. At the end of the baseline establishment period, the controller device can determine baseline values for different parameters in the telemetry data (such as baseline temperature, baseline voltage, baseline current draw, etc.). Baseline values can be established for individual network devices or groups of network devices. For example, baseline values can be established for network devices from the same manufacturer, network devices of the same brand and / or model, network devices in the same common area of the data center, network devices configured with the same software (operating system, applications, etc.), or other groups. After the baseline values for their respective parameters are established, network management system 10 can continue to receive telemetry data from network device 14. The network management system 10 can compare the currently received telemetry data with baseline data and use a threshold determined according to the anomaly detection intent provided by the administrator 12 to determine whether the network device in the network device 14 is operating normally, and in this way determine one or more anomalies in the network 2.
[0063] Anomalies may include, for example, misconfiguration of network device 14, cabling problems, policy violations, unexpected traffic patterns or other loads, or hardware failures. A list of example anomalies, affected nodes, and their descriptions is provided below, but additional categories and types of anomalies are envisioned.
[0064]
[0065]
[0066] Detecting anomalies
[0067]
[0068]
[0069]
[0070]
[0071]
[0072] Because the network management system 10 continuously updates the network diagram 13 to reflect the actual, real-time state of the network 2, the network management system 10 can identify anomalies by comparing the network diagram 13 with the intent 7 of the network 2. Any deviation between the network diagram 13 and the intent 7 indicates an anomaly.
[0073] When an anomaly is detected, the network management system 10 associates the anomaly with a specific node, edge, and / or one or more attributes in the network graph 13. For example, if network device 14B malfunctions or is misconfigured, the anomaly will be linked to a graph node in the network graph 13 representing network device 14B. As another example, if a link and / or cable-related anomaly exists, the network management system 10 associates the anomaly with an edge connecting the two nodes. The network management system 10 may output a user interface for display that depicts the network graph 13 and visually indicates the anomaly at the associated node or edge.
[0074] Network management system 10 can use graph queries to detect, analyze, and / or report anomalies. Such graph queries can traverse the graph to find discrepancies between the actual state of network 2 and the intended state 7. For example, queries can be written by the administrator 12 or the network management system 10 vendor and can be designed to examine the status and / or configuration of network devices 14 and other components of network 2, find missing or incorrectly configured paths (e.g., interfaces not assigned the correct VLANs), or analyze dependencies and their impact (e.g., how a failure of one device might affect the rest of the network), and for other purposes.
[0075] According to the technology disclosed herein, network management system 10 identifies multiple anomalies in network 2 that deviate from the intended network 2. These anomalies are each associated with one or more nodes in the network graph, for example, as attributes or "labels" of the nodes. The network graph augmented using the anomaly data can be a modified intention 7 or network graph 13. The anomaly data identifies anomalies. For example, anomaly data may indicate link failure, Link Layer Discovery Protocol (LLDP) neighbor error, BGP failure, LLDP missing, cable anomaly, BGP mismatch, excessive resource utilization, etc. In some cases, analysis system 17, or simply "system" 17, obtains a structured but non-graph form of intention from network management system 10 and processes the intention to generate a queryable intention network graph. Analysis system 17 is a computing system and can be integrated within the network management system or implemented and deployed to another computing system. Analysis system 17 applies predefined graph queries to the augmented network graph, which matches nodes with anomalies, relationships between these nodes, and the anomalies themselves. The predefined graph queries are associated with data indicating which of the anomalies matched by the predefined graph queries is the critical anomaly. Data can further indicate other anomalies associated with the critical anomaly. A critical anomaly is an anomaly that is, for example, the cause of other anomalies associated with the critical anomaly, the anomaly with the greatest impact among those matched by a predefined graph query, or an anomaly that is otherwise considered important (i.e., "critical") by the operator or expert. The analysis system can execute predefined graph queries about network graph 13 and, when a match is found with a node in network graph 13, output an indication that network 2 is experiencing a critical anomaly. The match can be a subgraph of network graph 13. Predefined graph queries and data can be generated, stored, and displayed as knowledge cards.
[0076] Graph queries can be configured to match specific nodes in network graph 13 by node identifier. Graph queries can also, or alternatively, be configured to match node types in network graph 13 by type (e.g., "system", "leaf", "trunk"). The latter case can effectively be a template that matches these subgraphs if many different subgraphs of network graph 13 also experience anomalies in matching graph queries. For example, a network may have many leaf-trunk pairings / links. Graph queries that match leaf types to trunk types will match these pairings.
[0077] In some technical aspects, the analysis system 17 can map critical anomalies present in network 2 to application-level problems. Services executing on server 11 (also referred to as "compute nodes") connected via network 2 and clients interacting with those services may be affected by critical anomalies. The analysis system 17 can store service impact data indicating one or more services that may be affected by critical anomalies. For example, a faulty interface can prevent access to services running on compute nodes that communicate using that interface. When identifying critical anomalies, the analysis system 17 uses service impact data to identify one or more services that may be affected by the critical anomalies. The analysis system 17 can output indications of one or more services. In some examples, the analysis system 17 extends a visual depiction of at least a portion of the network topology to include service topology, effectively extending network graph 13 to visually indicate services and / or clients affected by critical anomalies. Returning to the example above, the service impact data associated with a critical anomaly specifies that a faulty interface of a leaf switch may affect all services running on compute nodes connected to the faulty interface of the leaf switch. The analysis system 17 can extend the network topology to indicate services running on affected compute nodes and, in some cases, clients connected to those services. These indications of the affected services and clients can be considered synthetic anomalies because they were not identified by the network management system 10 using telemetry, configuration, or operational data from the network or computing nodes, but were identified as potentially caused by critical anomalies based on service impact data.
[0078] In some technical aspects, the network management system 10 associates operational data with one or more nodes in a network graph representing an intent (e.g., intent 7). For example, such operational data could indicate a faulty interface, a suspended / cold interface, interface oscillation, poor fiber optic cable, latency issues, resource utilization, environmental factors (fans, power, temperature), device traffic, configuration deviations, the number of routes in an EVPN, the size of an EVPN flood list, etc. As an example, the CPU utilization for a leaf switch could be 80%, the link could have a latency >1 ms, or the EVPN flood list could have 25 interfaces. Operational data is associated with one or more nodes in the network graph, for example, as attributes or "labels" of the nodes. The analysis system 17 applies predefined graph queries to the network graph, matching nodes with attributes that satisfy thresholds defined in the graph query, the relationships between those nodes, and the attributes that satisfy the thresholds. Predefined graph queries can be based on the network's Service Level Agreement (SLA) / Service Level Expectation (SLE). The predefined graph queries are associated with data that the analysis system 17 uses to identify and indicate one or more affected nodes in the network. Analysis system 17 can perform predefined graph queries against the network graph and, when a match is found with a node in the network graph, output an indication that one or more affected nodes in the network are experiencing a problem (e.g., poor health) or a positive indication that one or more "affected nodes" meet SLA / SLE requirements. Matching can represent subgraphs of the network graph that represent the intent.
[0079] In the above techniques, predefined graph queries and data can be generated, stored, and displayed as one of the knowledge cards 204.
[0080] Analysis system 17, network management system 10, and / or administrator 12 can operate to resolve one or more anomalies based on the determination of a matching subgraph indicating multiple anomaly associations and / or based on the identification of critical anomalies. For example, analysis system 17 can send indications of critical anomalies and / or indications of associations between multiple anomalies to network management system 10, causing network management system 10 to perform one or more actions to resolve at least one of the multiple anomalies. In some cases, this action is determined by, as follows: Figure 2 The action card discussed is specified. In some cases, the analysis system 17 can directly and automatically resolve at least one of multiple anomalies. In some cases, an operator (e.g., administrator 12) makes physical changes (e.g., rewiring), configuration changes, or other changes to network 2 to resolve at least one of multiple anomalies. In some cases, these operations and changes may be automatically performed by the analysis system 17 and / or the network management system 10 or in response to user input from administrator 12.
[0081] Figure 2This illustrates an analysis system 17 and a method for using the technology according to this disclosure. Figure 1 A block diagram of a set of example components of the network management system 10.
[0082] The network management system 10 and analysis system 17 may include processing circuitry 25, memory 27, one or more input devices, one or more communication units, and one or more output devices. Figure 2 Only processing circuitry 25 and memory 27 for network management system 10 are shown in this specification; however, in some embodiments, analysis system 17 may include separate instances of processing circuitry 25 and memory 27. In some examples, processing circuitry 25 includes one or more of a microprocessor, controller, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or equivalent discrete or integrated logic circuitry, or other types of processing circuitry 25. Network management system 10 and analysis system 17 may use processing circuitry 25 to perform one or more operations according to one or more aspects of this disclosure using software, hardware, firmware, or a combination of hardware, software, and firmware residing in and / or executing network management system 10 and analysis system 17, and may be distributed across one or more devices. One or more storage devices of memory 27 may be distributed across one or more devices. Processing circuitry 25 and memory 27 may provide an operating environment or platform for one or more modules or units, which may be implemented as software, but in some examples may include any combination of hardware, firmware, and software. Processing circuitry 25 is executable, and one or more storage devices (e.g., memory 27) may store instructions and / or data of one or more modules or units. The combination of processing circuitry 25 and memory 27 can retrieve, store, and / or execute instructions and / or data of one or more application programs, modules, units, or software. Processing circuitry 25 and / or memory 27 can also be operatively coupled to one or more other software and / or hardware components, including but not limited to… Figure 2 One or more components are shown.
[0083] In another example, the network management system 10 and the analysis system 17 are implemented on any suitable computing system having one or more computing devices, such as desktop computers, laptops, game consoles, smart TVs, handheld devices, tablets, mobile phones, smartphones, etc. In some examples, at least a portion of the network management system is distributed across cloud computing systems, data centers, or across networks (such as the Internet, another public or private communication network (e.g., broadband, cellular, Wi-Fi, and / or other types of communication networks)) for transferring data between computing systems, servers, and computing devices.
[0084] In some examples, the network management system 10 and the analysis system 17 are connected and communicate via a network. In some examples, the analysis system 17 is implemented as one or more modules of the network management system 10.
[0085] In this example, the network management system 10 includes a control unit 22, a network interface 34, and a user interface 36. The network interface 34 indicates that the network management system 10 can be communicatively connected to external devices (e.g., Figure 1 An exemplary interface of one of the network devices 14. Figure 2 Only network device 14A is shown in the diagram. Network interface 34 may represent a wireless and / or wired interface, such as an Ethernet interface or a wireless radio, configured to communicate according to one or more wireless standards, such as IEEE 802.11 wireless networking protocols (e.g., 802.11a / b / g / n or other such wireless protocols). In various examples, network management system 10 may include multiple network interfaces, although only one network interface is shown for illustrative purposes.
[0086] Control unit 22 refers to any combination of hardware, software, and / or firmware used to implement the functions of control unit 22 and its constituent modules and elements. When control unit 22 includes software or firmware, control unit 22 also includes any necessary hardware for storing and executing the software or firmware, such as one or more processors or processing units. Typically, processing units may include processing circuitry, one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or any other equivalent integrated or discrete logic circuitry, and any combination of these components. Furthermore, processing units are typically implemented using fixed and / or programmable logic circuitry.
[0087] User interface 36 represents one or more interfaces, such as user (e.g., administrator 12) Figure 1 The administrator 12 interacts with the network management system 10 through one or more interfaces, for example, to provide input and receive output. For example, the user interface 36 may represent one or more of a monitor, keyboard, mouse, touchscreen, touchpad, trackpad, speaker, camera, microphone, etc. Furthermore, although the network management system 10 includes a user interface in this example, it should be understood that the administrator 12 does not need to interact directly with the network management system 10, but can access the network management system 10 remotely, for example, via network interface 34.
[0088] In this example, control unit 22 includes a user interface module 38, a network interface module 32, a data collection module 37, and a management module 24. Control unit 22 executes user interface module 38 to receive input from user interface 36 and / or provide output to user interface 36. Control unit 22 also executes network interface module 32 to send and receive data (e.g., data packets) via network interface 34. User interface module 38, network interface module 32, data collection module 37, and management module 24 can again be implemented as corresponding hardware units, or as software, firmware, or a combination of both.
[0089] exist Figure 5 , Figures 6A to 6D , Figure 11 , Figure 12A and Figure 13A An exemplary user interface generated and output by the user interface module 38 or a similar user interface module of the analysis system 17 is depicted.
[0090] Control unit 22 can execute data collection module 37 to collect data from network devices (e.g., network device 14). Figure 1 Acquiring telemetry data. The data collection module 37 can store telemetry data in a telemetry database (DB) 39 as a time series of telemetry data. The data collection module 37 can acquire telemetry data from network devices using a "push" model or a "pull" model. In the push model, the network device (e.g., an agent on the network device) is configured to periodically send telemetry data to the data collection module 37. In the pull model, the data collection module 37 periodically requests the network device (e.g., an agent on the network device) to provide telemetry data to the data collection module 37. The service interval can be configured according to the type of telemetry data being collected. As an example, data can be collected every five seconds for an optical transceiver. The data collection module 37 can store telemetry data acquired during the baseline establishment period as historical telemetry data 41. In addition to storing telemetry data, the data collection module 37 can store timestamps associated with the telemetry data to indicate the time when the telemetry data was collected.
[0091] Control unit 22 executes management module 24 to manage various network devices, such as, Figure 1 Network device 14. Management includes, for example, based on data from the user (e.g., Figure 1 The administrator 12) receives instructions to configure and analyze network devices, and provides users with the ability to submit instructions to configure and analyze network devices. In this example, the management module 24 also includes a configuration module 26, a conversion module 28, an analysis module 29, and an anomaly detection module 31.
[0092] Management module 24 is configured to receive intents (e.g., high-level configuration instructions or anomaly detection instructions) from users (such as administrator 12 or another system (hereinafter referred to as "users")) for a set of managed network devices. In some examples, management module 24 may be referred to herein as a "structure manager". Over time, users can update configuration instructions, for example, to add new services, remove existing services, or modify existing services performed by managed devices. Furthermore, users can update anomaly detection instructions over time to change how analysis module 29 uses telemetry data to detect anomalies. Intents can be constructed based on, for example, YANG. In some examples, management module 24 also provides users with the ability to submit a conversion function, which conversion module 28 performs to convert the intent into device-specific low-level configuration instructions, as discussed below.
[0093] The network management system 10 also includes a configuration database 40. The configuration database 40 may include data structures describing the managed network devices (e.g., network device 14). The configuration database 40 may act as an intent data store, which can be used to store and manage a collection of intent data models. For example, the configuration database 40 may include information indicating device identifiers (e.g., MAC and / or IP addresses), device type, device vendor, device category (e.g., router, switch, bridge, hub, etc.). The configuration database 40 may store the current configuration information (e.g., intent data models, or in some cases, both intent data models and low-level configuration information) of the managed device (e.g., network device 14). The configuration database 40 may include a database containing intent data models. The configuration database 40 may be a graph database (graph DB), designed to represent and query data structured as a graph consisting of nodes, edges, and attributes.
[0094] Management module 24 can maintain data structures in configuration database 40. The data structures can include multiple vertices and multiple edges, where each vertex represents a corresponding network device (e.g., network device 14) or a corresponding stateless intent among multiple stateless intents, and the edges define relationships between the vertices. Management module 24 can receive indications of stateful intents. For example, management module 24 can receive configuration data from a user (such as administrator 12) for a unified graph model of intents for a set of managed network devices. These intents can be transformed and configured as graph data structures.
[0095] The conversion module 28 (which may also be referred to herein as the "device manager") can use the configuration database 40 to determine which devices it manages. Based on the information in the configuration database 40, the conversion module 28 determines which conversion functions 30 are executed on high-level configuration instructions, such as which devices will receive low-level configuration instructions (e.g., device-level configuration instructions). The conversion module 28 then executes each determined conversion function 30, providing the high-level configuration instructions as input and receiving the low-level configuration instructions. The conversion module 28 can then provide the low-level configuration instructions to the configuration module 26.
[0096] After receiving low-level configuration instructions from conversion module 28, configuration module 26 sends the low-level configuration instructions to the appropriate managed network device whose configuration needs to be updated via network interface module 32. Network interface module 32 passes the low-level configuration instructions to network interface 34. Network interface 34 forwards the low-level configuration instructions to the network device. In some examples, the function of conversion module 28 can be performed by the network device. For example, control unit 22 can output instructions for high-level configuration to the network device, and the agent of conversion module 28 operating at the network device can translate the received high-level configuration instructions into low-level configuration instructions for the network device.
[0097] Although it is for the purpose of allowing administrator 12 ( Figure 1 The example of interaction with network management system 10 is intended to describe user interface 36, but other interfaces may be used in other examples. For example, network management system 10 may include a representative state transfer (REST) client (not shown) that can act as an interface to another device through which administrator 12 can configure network management system 10. Similarly, administrator 12 can configure network device 14 by interacting with network management system 10 via a REST client.
[0098] Analysis module 29 can analyze telemetry data in telemetry database 39 to determine baseline data 42. For example, analysis module 29 can analyze the time series of data collected by data collection module 37 and stored as historical telemetry data 41 to determine the baseline operating characteristics of network devices, such as temperature, voltage, and current draw. Analysis module 29 can determine multiple sets of baseline data. For example, analysis module 29 can analyze the time series of data to determine the baseline operating characteristics of a specific network device and / or a group of network devices. For example, analysis module 29 can determine the baseline operating characteristics of a group of network devices from the same supplier, the same brand and / or model, located in the same location, etc. In some examples, analysis module 29 can determine the baseline operating characteristics regarding the time of day, the day of the week, the week of the year, etc. As an example, network devices (e.g., Figure 1Network device 14A can transmit more data during weekday working hours compared to off-peak hours and weekends. Therefore, baseline operating temperature, voltage, and / or current parameters can be higher during working hours than during off-peak hours. As an additional example, a data center can have different temperature characteristics in different parts of the data center. For example, a data center can have different cooling capacities in different areas of the data center, or some areas of the data center may have more equipment generating heat. Therefore, network devices in one area of the data center can have different baseline operating temperatures than network devices in other areas of the data center.
[0099] In some respects, baseline data 42 may be based on a time series of data collected over a thirty-day period from historical telemetry data 41. However, it may be other periods, longer or shorter than 30 days. Typically, the collection period may depend on the data storage capacity of the network management system 10. When new data is collected, analysis module 29 may use the new data to recalculate baseline data 42. For example, analysis module 29 may maintain baseline operating characteristics (such as baseline operating temperature or baseline voltage) as a moving average over the most recent thirty-day period.
[0100] Analysis module 29 can determine different parameters based on historical telemetry data 41 to generate baseline data 42 for network devices and network device groups. As an example, analysis module 29 can perform statistical analysis to determine different baseline statistical measurements associated with time series values of operating temperature, voltage, current draw, etc. For example, analysis module 29 can determine the mean, standard deviation, quantiles, percentile thresholds, probability density functions, etc. Analysis module 29 can use baseline statistics to determine outlier thresholds for various parameters associated with network devices and network device groups. Thresholds can set lower and / or upper limits for operating characteristics. Analysis module 29 can also perform regression analysis on time series data to determine relationships between operating characteristics and trends in the values of operating characteristics.
[0101] In some aspects, parameters may be based on individual operating characteristics, such as temperature, voltage, current draw, etc. In other aspects, parameters may be based on combinations of operating characteristics of network devices. Analysis module 29 can assign scores based on the values of combinations of operating characteristics. Furthermore, analysis module 29 can perform statistical analysis on scores determined based on time series of historical data. For example, analysis module 29 can determine the score for each set of telemetry data collected for a network device over time. Then, analysis module 29 can determine the mean, standard deviation, quantiles, percentile thresholds, probability density function, etc., of that set of scores. Analysis module 29 can use baseline statistics to determine outlier thresholds for scores related to network devices and groups of network devices. Analysis module 29 can also perform regression analysis on the time series of scores to determine the relationship between operating characteristics and scores, as well as the trend of score values.
[0102] The anomaly detection module 31 receives current telemetry data from the data collection module 37 and compares the current telemetry data with the threshold in the anomaly threshold 43. If the instantaneous (e.g., most recently acquired) value of a parameter determined based on the operational characteristics and / or network performance data in the telemetry data of the network device does not meet the anomaly threshold of the operational characteristics, the anomaly detection module 31 can determine that an anomaly event has occurred regarding the network device. The anomaly detection module 31 can store data related to the anomaly event in the telemetry database 39. The event-related data may include the timestamp of the event and the event type (overvoltage, undervoltage, overcurrent, undercurrent, overheating, etc.). The anomaly detection module 31 can generate an alarm indicating that an anomaly event has occurred. In some aspects, in response to an alarm, the anomaly detection unit can output details about the alarm on a network anomaly report. In some aspects, in response to an alarm, the administrator 12 can request the anomaly detection module 31 to generate user interface data 33 to present information about the alarm event. The anomaly detection module 31 can use the timestamp of the alarm event to obtain the telemetry data of the network device from the telemetry database 39. The anomaly detection module 31 can acquire telemetry data of the network device in three time periods: a first time period before an anomaly is detected, a second time period when an anomaly is detected, and a third time period after an anomaly is detected. The time periods can be set to default values, or the administrator 12 can specify the time periods to use. The anomaly detection module 31 can present baseline values of operational characteristics other than the values that caused the anomaly to be detected. For example, the anomaly detection module 31 can present baseline values of network device characteristics or the group to which the device belongs, and can also present the values that caused the anomaly to be detected. Additionally, the anomaly detection unit can present network traffic data flowing through the network device when an anomaly occurs.
[0103] The anomaly detection module 31 can perform linear regression on the time series database to determine whether the operating characteristics of the network device tend to deviate from the baseline value. If the rate of change exceeds a threshold, the anomaly detection module 31 can indicate an anomaly in the network device exhibiting this trend.
[0104] The anomaly detection module 31 can store abnormal data in the memory 27 or, for example, in an internal or external database, and can output the abnormal data 202 to the analysis system 17.
[0105] According to the technology of this disclosure, the analysis system 17 stores knowledge cards 204. Each knowledge card in the knowledge card 204 defines a method for identifying critical anomalies and their associated anomalies. A knowledge card is a data set containing (or including queries or other mechanisms for identification) specific patterns of nodes and edges in a network graph, as well as anomalies, health, or other attributes associated with the nodes in the network graph. The knowledge card also contains indications of critical anomalies and associated anomalies. A critical anomaly is an associated anomaly detected using a knowledge card graph query, and associated anomalies are a list of anomalies that may be caused by the critical anomaly. The knowledge card may also contain one or more of the following: a unique identifier for the knowledge card, a graph query language identifier, a version of the knowledge card indicating revisions, an organization identifier, a modification timestamp, an author, or an activity flag indicating whether the knowledge card is used for impact analysis. A user or organization will select one or more of the knowledge cards 204 and may set an activity flag to true to cause the analysis system 17 to use the selected knowledge cards for impact analysis.
[0106] Users can define new knowledge cards 204 using the user interface of the analysis system 17 or by providing data defining knowledge cards 204 via an interface (e.g., a REST interface). Graph queries used for knowledge cards can be the union of any subset of queries for pattern or symptom anomalies, and this union is mapped to a critical anomaly in the knowledge card. For example, cable, interface, configuration, and service anomalies can have associated graph queries, and the union of such graph queries can be set as a graph query for the knowledge card and mapped to cable cut as a critical anomaly. Users can be experienced expert users who understand the relationships between different anomalies and can associate critical anomalies due to their understanding of causal relationships within the network.
[0107] In the following example of knowledge card graph queries and anomalies, "Link broken" is the critical anomaly, and its associated anomalies are "Operation down," "LLDP missing," and "BGP." Therefore, a "Link broken" knowledge card can be created to identify a situation from the network graph where two interfaces experience operational failure, LLDP is missing at both ends of a link, and the BGP peer across that link experiences operational failure. This situation can be expressed using the following pseudograph query language to define the graph query for the knowledge card:
[0108] match(
[0109] node('system',name='sys_one',tags=not_none())
[0110] .where(lambda sys_one:'cabling_anomaly'and'link_broken'in sys_one.tags)
[0111] .out('hosted_interfaces',name='e_1').node('interface',name='int1',tags=not_none())
[0112] .where(lambda int1:'cabling_anomaly'and'bgp_mismatch'in int1.tags).out('link',
[0113] name='e2').node('link',name='link1',tags=not_none()).where(lambdalink1:
[0114] 'cabling_anomaly'in link1.tags) )
[0116] In some examples, the analysis system 17 stores action cards that it uses to analyze factors contributing to anomalies (whether critical or related). For example, given an ECMP imbalance, the analysis system 17 can execute actions on the action cards to identify the cause of the ECMP imbalance. For instance, actions on the action cards might prompt the analysis system 17 to perform actions to identify elephant flows in the path, poor hash quality, or missing routes.
[0117] In some examples, knowledge cards specify synthetic or anticipated anomalies. These are anomalies that are not determined by network management system 10 to have occurred, but rather are anomalies that may occur if, for example, analysis system 17 identifies a critical anomaly, they match a graph query on the knowledge card. Synthetic or anticipated anomalies allow analysis system 17 to associate (or “label”) nodes of the network graph (e.g., intent 7 or network graph 13), services, or clients with synthetic or anticipated anomalies. These anomalies can be used to predict or determine the potential impact on other nodes, services, or clients running on network 2 and to provide appropriate indications to the user. Because services and clients are not native to the intent network graph, analysis system 17 can also add nodes to the intent network graph representing services or clients to associate these nodes with synthetic or anticipated anomalies. As an example, a critical anomaly of missing routes would affect a virtual network. Knowledge cards can specify synthetic anomalies of virtual networks associated with routes in intents, network graph 13, or other network configuration or operational data. Analysis system 17 can then associate these synthetic anomalies of virtual networks as anomalies, even if the anomaly is synthetic because network management system 10 did not detect the anomaly in network 2. The analysis system 17 can output an indication of this synthetic anomaly to the user.
[0118] The user can select which of the knowledge cards 204 are active, i.e., used by the analysis system 17 when identifying related anomalies of key anomalies. The analysis system 17 may apply one or more of the knowledge cards 204, for example, on demand, periodically (e.g., every 1 second, every 5 seconds, every 30 seconds, etc.), or in response to receiving anomaly data 202 indicating a new anomaly.
[0119] In some examples, one or more modules of the network management system 10 may be implemented as part of the analysis system 17. For example, the anomaly detection module 31 may be implemented as part of the analysis system 17.
[0120] Figure 3A and Figure 3B This is a conceptual diagram illustrating an exemplary network device communicating with a network management system according to the technology of this disclosure. Figure 3A and Figure 3B For illustrative purposes only Figures 1 to 2 The discussion takes place against this background. Figure 3A and Figure 3B Network devices 314A and 314B can be Figure 1 The implementation method of network device 14. In Figure 3A and Figure 3BIn the example, network device 314A includes transceiver 304A and network device 314B includes transceiver 304B. In some aspects, transceivers 304A and 304B may be optical transceivers; however, this disclosure is not limited to such transceivers. Network devices 314A and 314B may include sensors 303A and 303B, respectively.
[0121] exist Figure 3A In the example, data collection module 37 may utilize an intent provided by administrator 12 to determine the telemetry data to be collected from network devices 314A and 314B. Based on the intent in response to determining that telemetry data should be collected from network devices 314A and 314B, data collection module 37 may initiate probes 301A and 301B. Probe 301 is configured to acquire telemetry data from the network devices. For example, probe 301A may be configured to acquire telemetry data from network device 314A using an application programming interface (API) or other interfaces provided by network device 314A, and probe 301B may be configured to acquire telemetry data from network device 314B using an API or other interfaces provided by network device 314B. In some aspects, the API or other interfaces used by the probes to collect telemetry data may be proprietary to the network devices. As an example, many network devices implement an "indicate" command that can be used by probes 301A and / or 301B to acquire telemetry data from such network devices. In some respects, the detector can use standardized interfaces such as SNMP to collect telemetry data.
[0122] The detector can send a request to the network device, indicating requested telemetry data. As an example, in response to receiving a request for telemetry data from detector 301A, network device 314A can acquire the requested telemetry data from sensor 303A and / or from transceiver 304A. Similarly, in response to receiving a request for telemetry data from detector 301B, network device 314B can acquire the requested telemetry data from sensor 303B and / or from transceiver 304B. Sensors (such as sensor 303A or 303B) can be configured to provide temperature data, current data, voltage data, etc. Although one sensor 303 and one transceiver 304 are shown for network devices 314A and 314B, network device 314 can have multiple sensors 303 and / or multiple transceivers 304. After acquiring their respective telemetry data, detectors 301A and 301B can provide their respective telemetry data to data collection module 37, which can store the telemetry data along with a timestamp in telemetry database 39 to indicate the time when the telemetry data was collected.
[0123] exist Figure 3BIn the example shown, the data collection module 37 can use the intent provided by the administrator 12 to determine the telemetry data to be collected from network devices 314A and 314B. In some aspects, the network management system 10 can transmit the type of telemetry data to be collected from network devices 314A and 314B. As an example, the network management system 10 can transmit a first set of telemetry collection parameters to agent 302A and a second set of telemetry parameters to agent 302B, informing agents 302A and 302B that they will collect operating temperature, operating voltage, operating current, etc., from their respective network devices 314A and 314B. Agents 302A and 302B can collect the indicated telemetry data and provide it to the data collection module 37 for storage as time-series data in the telemetry database 39. In some aspects, a push model can be used, in which agents 302A and 302B automatically and periodically provide their respective telemetry data to the data collection module 37. In some aspects, a request model can be used, in which agents 302A and 302B provide their respective telemetry data to data collection module 37 in response to a request received from data collection module 37.
[0124] Agents 302A and 302B can acquire operational characteristics included in telemetry data from various sources. As an example, agent 302A can acquire operational characteristics from sensor 303A and / or transceiver 304A. Similarly, agent 302B can acquire operational characteristics from sensor 303B and / or transceiver 304B. After acquiring their respective telemetry data, agents 302A and 302B can provide the telemetry data to data collection module 37, which can store the telemetry data along with a timestamp in telemetry database 39 to indicate the time the telemetry data was collected.
[0125] Figure 4 This illustrates the technology according to the present disclosure for use with Figure 1 The network diagram 700 is a conceptual diagram of the network. Network diagram 700 can be an intentional network diagram (e.g., intention 7) or a model of network states (e.g., network diagram 13). Figure 4 In the diagram shown, nodes 714A to 714G correspond to Figure 1 Network devices 14A to 14G. Node 714A, corresponding to network device 14A, is the root of the graph. From Figure 4As shown in the example, node 714B has five downstream nodes 714C to 714G. Node 714F has a single downstream node 714G. Therefore, a failure at node 714B affects more nodes than a failure at node 714F. The anomaly detection module 31 can use the number of affected nodes to determine the risk factors associated with the anomaly. Therefore, in this example, an anomaly at network device 14B (represented by node 714B) poses a higher risk than an anomaly at network device 14F (represented by node 714F). In some aspects, the anomaly detection module 31 can generate a graphical representation of the network graph in which nodes experiencing anomalies are highlighted. In some aspects, the graphical representation can differ from nodes posing a lower risk of failure by highlighting nodes posing a higher risk (e.g., color coding).
[0126] Figure 5 An exemplary user interface for displaying a network graph according to the technology of this disclosure is depicted. Network graph 578 is a directed acyclic graph. The user can provide user input to the user interface to interact with the network graph. As shown, the user has selected nodes 580A to 580D (collectively referred to as "node 580"), which has the following effects: (1) highlighting the nodes in the selected node 580 and the edges that come into or out of it, and (2) displaying a corresponding pop-up window showing the properties of node 580.
[0127] The types of nodes used in the network graph in intent-based networks can include the following:
[0128] Device or system node refers to a physical device in a network, such as a switch (e.g., a backbone switch and leaf switch in a Clos architecture), a router, a server (e.g., a storage or compute server), a firewall, a load balancer, or a storage device.
[0129] An interface node represents a single network interface or port on a device, such as an Ethernet port or a logical interface (e.g., VLAN, LAG).
[0130] Logical nodes represent abstract or logical entities in a network, such as Virtual Local Area Networks (VLANs), Virtual Routers and Forwarders Instances (VRFs), routing protocols (e.g., BGP, Shortest Path First Open-OSPF), or IP subnets.
[0131] Link nodes represent physical or logical links between devices, such as cable connections (physical links), overlay / virtual network connections (logical links), and link aggregation groups (LAGs).
[0132] A service node refers to a service running on the network, such as Dynamic Host Configuration Protocol (DHCP), Domain Name System (DNS), or IP Address Management (IPAM).
[0133] Policy nodes represent security or operational policies applied to the network, such as access control lists (ACLs), firewall rules, and quality of service (QoS) policies.
[0134] Group or role node, representing a group or role of devices in the network, such as device roles (e.g., "trunk", "leaf", "boundary leaf") or rack groups (e.g., devices in the same rack).
[0135] The types of edge relationships between nodes in a network graph used in intent-based networks can include the following:
[0136] Connection relationships refer to the physical or logical connections between devices, interfaces, or links, such as the connection between a leaf switch and a core switch, the relationship between a server and its connected leaf switches, or the link aggregation connection (LAG) between two devices.
[0137] Routing relationships refer to the relationships formed by routing protocols regarding how data packets are forwarded in the network, such as BGP peering between two routers or OSPF adjacency between two devices.
[0138] Membership refers to the interfaces, devices, or logical entities included in a specific group or domain, such as interfaces that are members of a specific VLAN, device role assignments (e.g., nodes that are part of a "backbone" group), or VRF associations between devices and virtual routing instances.
[0139] Service relationships refer to the relationships between network entities and the services they support or provide, such as the relationship between a DHCP server and the subnets it serves, or the relationship between a DNS server and the device that uses it for name resolution.
[0140] Policy relationships indicate the application of policies or rules to a specific network device, interface, or group, such as access control lists (ACLs) applied to a specific interface, firewall rules applied to traffic between two VLANs, or QoS policies applied to prioritize certain types of traffic.
[0141] Traffic flow relationships represent the actual data flow paths through the network, helping to analyze traffic flow from one node to another.
[0142] An exemplary edge relationship and node subgraph of a network graph showing a server-to-switch connection is shown below. The subgraph has leaf switch nodes, server nodes, and interface nodes representing Ethernet ports on both the switch and the server. There are connections between the interface nodes of the switch and the interface nodes of the server, representing physical connections.
[0143] Figures 6A to 6DA user interface generated and output for display according to the technology of this disclosure is depicted. The user interface can be generated and output by an analysis system 17 that acquires the intent of the anomalous data 202 and the network 2. The intent can be in the form of a graph, which allows the anomalous data to be augmented at the nodes of the intent and presents a visualization of the graph with the anomalous data.
[0144] Figure 6A A user interface 600 depicts a simple data center topology 604 showing switches and hosts, and a list of anomalies 602 detected by the network management system 10. Devices marked with anomalies are visually associated with user elements indicating "anomaly exists." The user interface 600 does not display any critical anomalies; instead, it displays all detected anomalies using a list of anomalies 602. While the topology 604 is relatively simple, the list of anomalies 602 is extensive, and investigating the root cause of the anomalies would be time-consuming and challenging for the user.
[0145] Figure 6B User interface 620 is depicted, similar to user interface 600, but showing a complete list of key anomalies instead of anomaly 602. Analysis system 17 associates the anomalies indicated by anomaly data 202 with nodes of intent. Analysis system 17 applies knowledge cards 204 mapped to configuration anomalies as key anomalies, and more specifically, applies graph queries of the knowledge cards to a network graph representing the intent, which is augmented with anomalies as attributes of nodes. Based on matching the queries with network graph 13, analysis system 17 has identified configuration anomalies with two associated anomalies (“Bad Wiring” and “BGP Mismatch”). Analysis system 17 can therefore group the anomalies and generate user interface 620 to display the anomalies under user element 622 of configuration anomalies, where the user can expand to display details of associated anomalies by selecting drop-down menus; in this example, two affected services are shown. User interface 620 can instead present one or more key anomalies to the operator, avoiding the hundreds or even thousands of anomalies to review and investigate, thus preventing alert fatigue; once a key anomaly is investigated and fixed, those anomalies associated with it may also be fixed. This allows operators or another system to resolve network-related issues more quickly. The user interface 620 provides a comprehensive view of the problems and impacts of applications / services running on the network and facilitates the differentiation of which anomalies are side effects of critical anomalies or which are unrelated to critical anomalies.
[0146] Figure 6BA user interface 640 is depicted, showing a topology layout 644 representing topology 604. In some examples, the analysis system 17 maps critical anomalies present in the network to application-level issues. Services executing on compute nodes connected via the network (hosts 1 to 4, as shown) and clients interacting with these services may be affected by critical anomalies. The analysis system 17 stores service impact data indicating one or more services that may be affected by critical anomalies. For example, a faulty interface may prevent access to services running on compute nodes that communicate using the interface. Service impact data can identify one or more services executing on a specific server 11 in network 2. Using such service impact data, the analysis system 17 can determine whether server 11A is experiencing or affected by an anomaly associated with a critical anomaly (or the critical anomaly itself), and all services identified as executing on server 11A are affected. Service impact data can identify one or more services that may be affected by critical anomalies by identifying one or more services in a corresponding knowledge card (e.g., "telnet") to indicate that the identified services are susceptible to impact when executed on a server affected by a critical anomaly in some way. If the analysis system 17 determines that server 11A is experiencing or affected by an anomaly associated with a critical anomaly (or the critical anomaly itself), the identified services performed on server 11A will be affected.
[0147] When identifying critical anomalies, the analysis system 17 uses service impact data to identify one or more services that may be affected by the critical anomalies. For example... Figure 6B As described, the analysis system 17 can use service impact data to determine any service, or specifically designated service, that might be affected by a configuration anomaly of the leaf switch if it is running on any host connected to p-acs-0-leaf1. This includes hosts 1 through 4. The analysis system 17 can output indications of one or more of those identified services. The analysis system 17 generates a user interface 620 to list two affected services: http and telnet.
[0148] In some examples, the analysis system 17 extends the visual depiction of at least a portion of the network topology to include service topology, effectively expanding the network graph to visually indicate the services and / or clients affected by the critical anomaly. Service impact data associated with the critical anomaly can specify that the configuration anomaly may affect all services running on compute nodes connected to the affected leaf switch. For example... Figure 6CAs depicted, the analysis system 17 generates a user interface 640 to extend the topology of the simple network to indicate services 686A to 686B running on the affected hosts 1 and 4 (note that the hosts themselves are not shown as experiencing anomalies) and to indicate clients 688A to 688B connected to or otherwise communicating with services 686A to 686B. These indications of affected services and clients displayed on the user interface 640 can be considered synthetic anomalies because they are not identified by the network management system 10 using telemetry, configuration, or operational data from the network or hosts, but rather based on service impact data identified as potentially caused by critical anomalies.
[0149] like Figure 6D As depicted, users can interact with user interface 640 to filter affected devices, services, and / or clients within the displayed network topology. This allows users to "focus" on affected areas of the network, reducing the scope of investigation or review. In response to user input or configuration of the filtered affected devices, services, and / or clients, analysis system 17 generates user interface 641, which includes user elements of the devices exhibiting anomalies, those devices connected to those devices, and clients connected to the host through those devices.
[0150] Figure 7A A network graph 800 is described, having nodes V1 through V9 (collectively referred to as “nodes V”). The network graph 800 may represent an intent (e.g., a “blueprint”) or may represent the current state of the network according to a graph data model (e.g., network graph 13).
[0151] Figure 7B The text describes how nodes V, marked with anomalies A1 to A6 (collectively referred to as "anomalies A"), are updated. Figure 7A In network graph 800, anomalies A1 through A6 are caused by cable cuts or ECMP imbalances in the network. The updated network graph is network graph 802. For example, node V8 has been marked with anomalies A5 and A6. Analysis system 17 can obtain anomaly data 202 and network graph 800 from network management system 10. To mark node V with one or more anomalies indicated in anomaly data 202, analysis system 17 can identify the appropriate node for each anomaly and associate the anomaly with the appropriate node. The appropriate node can be a node experiencing or causing an anomaly. To associate anomalies with nodes, analysis system 17 can add key:value pairs for the anomalies, where the key is an indicator of the anomaly value and the value indicates the type of anomaly. For example, Figure 10The node data 910A shows "tags:cabling_anomaly", which associates anomalies of the "cable anomaly" type with the corresponding node. The analysis system 17 can perform graph queries on the updated network graph 802 to identify associated anomalies of the critical anomaly, such as those described in other detailed descriptions in this disclosure.
[0152] Figure 7C A network graph 802 depicts subgraphs 860A to 860B representing queries for two knowledge cards 850A to 850B (collectively referred to as "knowledge card 850"). Knowledge card 850 is a "cable cut" knowledge card and can be used to identify a critical anomaly when a subgraph with a labeled anomaly matches a graph query for knowledge card 850A, indicating that the critical anomaly is a cable cut. Subgraphs can be specified using a query language, but in... Figure 7C The diagram illustrates two paths: Path 1: V5(A1)→V2(A2)→V1→V7(A3), and Path 2: V5→V4→V8(A5), where VN(AX) matches the node VN marked with an anomaly AX. Subgraph 860A exists in network graph 802. Therefore, the critical anomaly of cable cut exists in the corresponding network, and other anomalies in the query are associated with this critical anomaly. For example, if A1 is the critical anomaly, then anomalies A2, A3, and A5 are associated with the critical anomaly. Matching subgraph 860B is also shown for graph queries in knowledge card 850B and can be similarly used to identify associated anomalies of critical anomalies of ECMP imbalance.
[0153] Figure 8 This is an example diagram query based on the technology of this disclosure. Diagram query 890 is designed to identify related anomalies of configuration anomalies, specifically "interface closed," where the interface's configuration data specifies that the interface is disabled. Such configuration triggers numerous additional anomalies, all of which are detected by the network management system 10 and output to the analysis system 17. Diagram query 890 can be associated with a list of critical anomalies and related anomalies. For example, "config_anomaly" can be a critical anomaly, along with related anomalies such as "cabling_anomaly" and "bgp_down" (two instances, one on each side of the link) for two interface nodes, and "lldp_missing" for a link node.
[0154] Figure 9 The technology according to this disclosure is described and Figure 8The graph query matches an example subgraph of the network graph. The subgraph includes interface (IF) nodes 902A and 902C connected to link node 902B, where link node 902B is a link used to connect the corresponding interface. Exceptions 904A, 904B, and 904C are associated with interface nodes 902A, 902B, and 902C to match graph query 890.
[0155] Figure 10 The attributes of nodes that have been augmented with anomaly data according to the technology disclosed herein are listed. Node attribute 910A is an attribute of interface node 902A and includes the structured key:value form labels “cabling_anomaly” and “bgp_down”, which indicate that these anomalies are associated with interface node 902A. Node attribute 910B is an attribute of link node 902B, and node attribute 910C is an attribute of interface node 902C, and these attributes are similarly augmented.
[0156] Figure 11 A user interface generated by an analysis system according to the technology of this disclosure is depicted. The user interface includes a query editor that presents a graph query 890. Users (e.g., expert users) can use the user interface to edit graph queries to create graph queries for identifying associated anomalies of key anomalies. Graph queries may be included in knowledge cards. The graph on the user interface shows a subgraph for the current graph query. The subgraph may be matched with the network graph of the network.
[0157] Figure 12A A user interface generated by an analysis system according to the technology of this disclosure is depicted. The user interface is similar to... Figure 11 The user interface is depicted in the image. The graph query within the user interface is a knowledge card for link breakage, used to identify situations where a link connecting two interfaces is broken. Figure 12B The attributes of nodes that have been augmented with anomalous data according to the technology of this disclosure are listed. These nodes marked as anomalous are associated with... Figure 12A The user interface uses graph query matching. A critical exception can be "link_break," while other related exceptions are identified through graph queries.
[0158] Figure 13A A user interface generated by an analysis system according to the technology of this disclosure is depicted. The user interface is similar to... Figure 11 The user interface is depicted in the image. The graph query within the user interface is a knowledge card for link misscables, used to identify situations where a link is connected to the wrong adjacent device. Figure 13B The attributes of nodes that have been augmented with anomalous data according to the technology of this disclosure are listed. These nodes marked as anomalous are associated with... Figure 13AGraph query matching in the user interface. The critical exception can be "link_miscaled", and other related exceptions are identified through graph queries.
[0159] Figure 14 This is an exemplary system that implements the analytics system 17 and network management system 10 in more detail according to the techniques of this disclosure. In this example deployment, the network management system 10 is a local (1410, on-prem) solution. The analytics system 17 can be implemented as a Software as a Service (SaaS) product, focusing on providing enhanced operational capabilities to solve complex data center problems performed in a cloud computing system 1420. Cloud storage 1440 represents a cloud storage system used by the analytics system 17 to implement the techniques of this disclosure. Event streaming 1442 represents a service used by the analytics system 17 and can implement data pipelines, event streaming / messaging for event publishing / subscription, data integration, logging, and monitoring.
[0160] The NMS operating system (OS) 1452 implements the functions of the network management system 10 described elsewhere in this disclosure. A flow collector 1450 collects and analyzes data center network traffic. The flow collector 1450 simplifies the collection of network traffic flows and telemetry by providing seamless integration with organization-specific information. The flow collector 1450 provides visibility and insight into network traffic by providing granular information about network traffic flows, congestion, high latency, and packet loss; enables strategies to optimize network traffic flows, ensuring the most efficient use of available resources; and improves security by more effectively detecting and responding to threats while maintaining regulatory compliance.
[0161] The NMS agent 1454 of the network management system 10 can output abnormal data 202, intent, and any other data required by the analysis system 17 to perform the techniques described in this disclosure.
[0162] Cloud entry point 1446 is a service running in cloud computing system 1420 and is the entry point for any connected device that allows any edge component to communicate with the cloud.
[0163] Figure 15This is an exemplary system that implements the analysis system 17 and the network management system 10 in more detail. In some examples, the analysis system 17 and the network management system 10 may be combined within the overall network management system. In this example, anomalies identified by the network management system 10 for intent are sent to an event stream 1442 via a cloud entry point 1446 connected to the network management system 10. These anomalies included in the data center event topic 1514 are enriched in the data pipeline processing module 1510 that implements the anomaly topology 1512, and are sent to a separate topic, namely the data center enriched event topic 1516, in the event stream 1442. The data pipeline processing module 1510 processes the data stream. In this example, the data pipeline processing module 1510 enables the establishment of a topology (i.e., a data pipeline) for processing the data stream. A topic is a named channel or category to which messages are published in publish / subscribe, lightweight messaging, message queues, distributed logs, or event streaming platforms, and subscribers receive messages from that topic.
[0164] Stream processor 1534 job – Exception stream processor job 1536 – periodically aggregates these expanded exceptions and stores the aggregated data in cloud storage 1440. Workflow orchestration module 1518 periodically schedules impact analysis job 1520, which processes the data and stores key exceptions, related exceptions, affected services, and clients for search and analysis system 1526. Analysis system 17 can query elastic searches of the data to generate user interfaces, such as those depicted in the figure and described herein. A job is a discrete, scheduled, or triggered operation that performs a specific function within a workflow. Jobs can be defined by code, configuration, or task templates.
[0165] Impact analysis job 1520 runs graph queries on graph database 1532 and uses flow analyzer module 1530 to analyze flows to identify or obtain affected services and clients. Graph generator 1528 generates a graph that is stored in graph database 1532. Graph generator 1528 can generate graphs based on information about the network. Network graph 13 is an example of a graph stored in graph database 1532. Knowledge cards (“KC”) are stored in database system 1522 through database system interface 1524 and can be retrieved from database system 1522 by impact analysis job 1520 through database system interface 1524. Each of database system 1522 and graph database 1532 allows the creation, management, access, and manipulation of structured data and may include a database and a database management system. Graph database 1532 can use a graph structure in which data is represented as nodes and edges (relationships between nodes) to store and manage data, and nodes and / or edges may each have one or more associated attributes. Attributes can be represented as key:value pairs.
[0166] Figure 16 This is a flowchart illustrating an exemplary operating mode of an analysis system according to the technology of this disclosure. The flowchart operation is a description of analysis system 17, but can be performed by a separate network management system or other systems consistent with the technology of this disclosure.
[0167] The system (e.g., analysis system 17) acquires a graph query and a network graph for network 2. The network graph includes one or more nodes with one or more attributes indicating multiple anomalies of network 2 that are inconsistent with the intent of network 2. The network graph may have other nodes with one or more other attributes indicating other anomalies of network 2. The network graph may be based on the intent of network 2. Analysis system 17 performs a graph query on the network graph to determine matching subgraphs of the network graph (1602). The graph query matches one or more nodes and one or more attributes of the one or more nodes. That is, the subgraph includes one or more nodes. If the graph query does not match a subgraph of the network graph (the "No" branch of 1604), analysis system 17 takes no action (1606). Knowledge cards can be used to specify the graph query.
[0168] However, based on the determination of the matching subgraphs of the network graph (the "yes" branch of 1604), the analysis system 17 outputs an indication of the correlation between multiple anomalies (1608). The indication of the correlation between multiple anomalies can be a visualization of at least the matching subgraphs, an indication of key anomalies, or a list of one or more of the multiple anomalies.
[0169] In some cases, based on indications of associations among multiple anomalies, network management system 10 may reconfigure the network to resolve at least one of the anomalies. In some cases, based on the determination of a matching subgraph, analysis system 17 may instruct network management system 10 to reconfigure the network to resolve at least one of the anomalies. This may include specifically resolving critical anomalies, which would favor resolving anomalies that may have been identified using graph queries as being associated with critical anomalies.
[0170] The techniques described herein can be implemented using hardware, software, firmware, or any combination thereof. The various components, functional units, and / or modules shown in the accompanying drawings and / or elsewhere in this disclosure can perform operations described using software, hardware, firmware, or a mixture of hardware, software, and firmware residing in and / or executing on one or more computing devices. For example, a computing device may use multiple processors or multiple devices to execute one or more of these modules. A computing device may execute one or more such modules, such as a virtual machine executing on underlying hardware. One or more of these modules may execute as one or more services of an operating system or computing platform. One or more of such modules may execute as one or more executable programs at the application layer of a computing platform. In other examples, the functionality provided by the modules may be implemented by dedicated hardware devices. Although certain modules, data stores, components, programs, executable files, data items, functional units, and / or other items included within one or more storage devices may be shown individually, one or more of these items may be combined and operate as a single module, component, program, executable file, data item, or functional unit. For example, one or more modules or data stores may be combined or partially combined such that they operate as a single module or provide functionality. Furthermore, one or more modules may operate in combination with each other, such that, for example, one module serves or extends another module. Moreover, each module, data store, component, program, executable file, data item, functional unit, or other item shown within the storage device may include multiple components, sub-components, modules, sub-modules, data stores, and / or other components or modules or data stores not shown. Furthermore, each module, data store, component, program, executable file, data item, functional unit, or other item shown within the storage device can be implemented in different ways. For example, each module, data store, component, program, executable file, data item, functional unit, or other item shown within the storage device may be implemented as part of an operating system executing on a computing device.
[0171] If implemented in hardware, this disclosure is applicable to a device, such as a processor or integrated circuit device, such as an integrated circuit chip or chipset. Alternatively or additionally, if implemented in software or firmware, the technology may be implemented at least in part by a computer-readable data storage medium comprising instructions that, when executed, cause a processor to perform one or more of the methods described above. For example, a computer-readable data storage medium may store such instructions that are executed by a processor.
[0172] Computer-readable media can form part of a computer program product, which may include packaging material. Computer-readable media may include computer data storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic or optical data storage media, etc. In some examples, the article of writing may include one or more computer-readable storage media.
[0173] In some examples, computer-readable storage media may include non-transitory media. The term "non-transitory" may mean that the storage medium is not embodied in a carrier wave or propagating signal. In some examples, non-transitory storage media may store data that may change over time (e.g., in RAM or cache).
[0174] The code or instructions can be software and / or firmware executed by processing circuitry including one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other equivalent integrated or discrete logic circuits. Therefore, the term "processor" as used herein can refer to any of the foregoing structures or any other structure suitable for implementing the techniques described herein. Additionally, in some aspects, the functionality described in this disclosure can be housed within a software module or a hardware module.
Claims
1. A system for identifying anomalies in a network, comprising: A memory stores graph queries for a network and a network graph, wherein the network graph includes one or more nodes, the one or more nodes having one or more attributes indicating multiple anomalies of the network; and One or more processors are connected to the memory. The memory stores instructions that, when executed, cause the one or more processors to: Perform the graph query on the network graph of the network to determine a matching subgraph of the network graph, wherein the graph query matches the one or more nodes and the one or more attributes; and Based on the determination of the matching subgraph, an indication of the association between the multiple anomalies is output.
2. The system according to claim 1, in, The memory stores the association between key anomalies and the graph query. The indications of the correlation between the multiple anomalies include the indications of the critical anomalies.
3. The system according to claim 1, in, The memory stores the associations between the plurality of anomalies and the graph query, wherein the indications of the associations between the plurality of anomalies include the indications of the plurality of anomalies.
4. The system according to claim 1, wherein, When the instruction is executed, it causes the one or more processors to: The interface receives knowledge cards that associate key exceptions with the graph query.
5. The system according to claim 4, in, The knowledge card also includes data indicating the plurality of anomalies, and Specifically, performing the graph query includes matching the plurality of anomalies with the plurality of attributes of the plurality of nodes in the network graph.
6. The system according to any one of claims 1 to 5, wherein, The indication of the correlation between the multiple anomalies includes at least the visualization of the matching subgraph.
7. The system according to any one of claims 1 to 5, in, The network graph includes a second network graph, and The memory stores instructions that, when executed, cause the one or more processors to: Based on the abnormal data indicating the multiple anomalies, the first network graph is modified to add the one or more attributes to the one or more nodes of the first network graph, thereby generating the second network graph.
8. The system according to claim 7, in, The memory stores instructions that, when executed, cause the one or more processors to: Receive the abnormal data from the network management system.
9. The system according to claim 7, wherein, The first network graph includes either an intentional network graph or a network graph that models the configuration and operational state of the network.
10. The system according to any one of claims 1 to 5, in, The memory stores the association between one of a synthetic anomaly and a predicted anomaly and the graph query, wherein the synthetic anomaly and the predicted anomaly indicate a potential impact on a service or client running on the network, and The memory stores instructions that, when executed, cause the one or more processors to: Based on the determination of the matching subgraph, an indication is output for one of the synthesized anomaly and the expected anomaly.
11. The system according to claim 10, wherein, Indications for either the synthesized anomaly or the anticipated anomaly include: visualization of at least the matching subgraph and additional nodes connected to at least the matching subgraph and representing the service or client.
12. The system according to any one of claims 1 to 5, wherein, The memory stores instructions that, when executed, cause the one or more processors to: Based on the determination of the matching subgraph, the network is reconfigured to resolve at least one of the plurality of anomalies.
13. The system according to any one of claims 1 to 5, wherein, The memory stores instructions that, when executed, cause the one or more processors to: Based on the determination of the matching subgraph, the network management system is guided to reconfigure the network to resolve at least one of the plurality of anomalies.
14. A method for identifying anomalies in a network, comprising: The system obtains a graph query and a network graph for the network, wherein the network graph includes one or more nodes, and the one or more nodes have one or more attributes that indicate multiple anomalies of the network; The system performs a graph query on the network graph of the network to determine a matching subgraph of the network graph, wherein the graph query matches one or more nodes and one or more attributes; and Based on the determination of the matching subgraph, an indication of the association between the multiple anomalies is output.
15. The method of claim 14, further comprising: Obtain the association between key anomalies and the graph query; as well as Based on the association between the key anomaly and the graph query, an indication of the association between the multiple anomalies is output to include an indication of the key anomaly.
16. The method of claim 14, further comprising: Obtain the association between the multiple anomalies and the graph query; as well as Based on the association between the multiple anomalies and the graph query, an indication of the association between the multiple anomalies is output to include an indication of the multiple anomalies.
17. The method according to any one of claims 14 to 16, further comprising: The graph query is associated with either a synthetic anomaly or a predicted anomaly, wherein the synthetic anomaly or the predicted anomaly indicates a potential impact on a service or client running on the network; and Based on the determination of the matching subgraph, an indication is output for one of the synthesized anomaly and the expected anomaly.
18. The method according to any one of claims 14 to 16, further comprising: Based on the determination of the matching subgraph, the network is reconfigured to resolve at least one of the plurality of anomalies.
19. The method according to any one of claims 14 to 16, further comprising: Based on the determination of the matching subgraph, the network management system is guided to reconfigure the network to resolve at least one of the plurality of anomalies.
20. A computer-readable storage medium encoded with instructions for configuring one or more programmable processors to a computing system according to any one of claims 1 to 13, or to perform the method according to any one of claims 14 to 19.
