A cloud computing virtual machine scheduling method suitable for railway communication cloud

Abstract

The application provides a cloud computing virtual machine scheduling algorithm suitable for railway communication cloud, aiming to solve the deficiency of existing scheduling algorithms in handling high real-time and high security requirements in railway communication systems. The algorithm performs accurate scheduling based on the model and bandwidth attributes of PCI network cards, ensuring that virtual machines can directly access physical network cards to reduce latency and improve data transmission rate. At the same time, the algorithm realizes network isolation between virtual machines through physical switch configuration, ensuring the security of the system. When creating virtual machines in batches, the scheduling algorithm can schedule physical server nodes that meet the topology requirements according to the user-defined network topology structure, ensuring the correct connection between virtual machines and network topology. Through this algorithm, the railway communication system can meet the requirements of low latency, high bandwidth, security isolation and batch scheduling in complex network environments, significantly improving the efficiency and flexibility of virtual machine scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cloud computing, and particularly to a cloud computing virtual machine scheduling method suitable for railway communication cloud. BACKGROUND

[0002] The railway communication system is the core technology to ensure the safety of train operation, and has extremely strict requirements on network performance. With the advancement of railway modernization, a large amount of real-time data generated during train operation must be processed and transmitted within a very short time, which directly affects the safety and efficiency of train operation. Therefore, the railway communication system has extremely high requirements on the low latency, high bandwidth, stability and reliability of the network. However, the traditional cloud computing scheduling technology cannot fully meet these strict performance requirements.

[0003] In the cloud computing environment, virtual machine scheduling technology is usually used to reasonably allocate server resources to improve efficiency, and through virtual network technology to ensure flexibility and scalability. However, there is a certain performance loss problem in virtual network, especially in the face of high real-time requirement scenarios, this performance loss may have a serious impact on train operation. For example, the increase of data transmission delay may cause the control command to be unable to be issued in time, affecting the safety of train operation.

[0004] In order to solve this delay problem, PCI pass-through technology is introduced in the cloud computing environment, which allows virtual machines to directly access physical network card resources, avoiding the performance loss of the virtualization layer. PCI pass-through technology directly maps physical network devices to virtual machines by bypassing the virtual network layer, greatly reducing the data transmission delay and improving the data transmission rate. However, although the PCI network card pass-through technology has a significant effect on performance improvement, it also brings new problems, especially in network isolation and security.

[0005] Traditional virtual network technology can achieve isolation and security protection between virtual machines through security groups, but PCI pass-through technology does not support the security group function in the virtual network, making the isolation problem between virtual machines more complex. In order to solve this problem, it must rely on the network isolation mechanism at the physical level, usually through the configuration of physical switches to achieve network isolation, which puts higher requirements on network architecture design. In the railway communication system, due to its complex network environment and high security requirements, the configuration of physical network not only needs to meet the isolation requirements, but also must ensure that it can quickly switch when the network fails, guaranteeing the continuity and high reliability of the network.

[0006] In addition, in the railway communication system, the deployment position of the virtual machine must be closely related to the network topology. This means that in the scheduling process of the virtual machine, in addition to considering the load balancing and availability of resources, the virtual machine needs to be accurately scheduled in position according to the specific network topology structure of the railway communication system. Through this scheduling, the virtual machine can be correctly connected to the specified physical network card and network segment, thereby ensuring the isolation, real-time performance and efficiency of the network.

[0007] The existing cloud computing scheduling algorithm mostly allocates virtual machines to suitable computing nodes based on resource requirements, load balancing and availability, and cannot meet the special requirements of the railway communication system on network performance and isolation. Especially in the face of a railway communication system with a complex network topology, the traditional scheduling algorithm cannot perform batch scheduling according to the connection requirements of the network topology when creating virtual machines in batches, nor can it flexibly handle the PCI card attributes when scheduling virtual machines. This makes it necessary to develop a customized virtual machine scheduling algorithm that can schedule according to the PCI device attributes and network topology structure to solve these problems.

[0008] Therefore, in view of the special requirements of the railway communication system, the present application proposes a cloud computing virtual machine scheduling algorithm suitable for railway communication clouds, which solves the deficiencies of existing scheduling algorithms in network performance, virtual machine isolation and batch scheduling, and ensures efficient virtual machine scheduling under the premise of safety and stability. SUMMARY

[0009] The present application aims to solve the problems raised in the background art by providing a cloud computing virtual machine scheduling method suitable for railway communication clouds.

[0010] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0011] A cloud computing virtual machine scheduling method suitable for railway communication clouds, the method comprising the following steps:

[0012] Step 1: Collect the hardware resource information of each physical server in the cloud platform, including CPU, memory, model of PCI network card, bandwidth, physical location and network connection topology, etc.

[0013] Step 2: Input and define the network topology structure of the railway communication system by the user, including network type, topology form, device type and network connection requirements, the defined network topology will determine the network connection and resource allocation between virtual machines;

[0014] Step 3: According to the user-defined network topology and the resource information of the physical server, a resource requirement list of the virtual machine is generated, which includes the required CPU, memory, PCI card bandwidth, network interface type, connection relationship between virtual machines, etc.

[0015] Step 4: According to the resource requirement list of the virtual machine, a scheduling algorithm is used to filter the physical server nodes that meet the requirements. The algorithm first filters according to the PCI card attributes (such as model, bandwidth) to ensure that the physical server meets the bandwidth and performance requirements of the virtual machine.

[0016] Step 5: After filtering the physical servers that meet the PCI card attribute requirements, further filtering is performed on the physical servers according to the network topology requirements of the virtual machine through a batch scheduling algorithm to ensure that the network connection between virtual machines meets the predefined topology structure.

[0017] Step 6: Batch create virtual machines on the physical server nodes selected by the above filtering steps, and establish correct network connections between each virtual machine and its corresponding physical network card and network connection topology to ensure that the virtual machine can run normally and meet the requirements of the railway communication system for network performance and isolation.

[0018] Step 7: After completing the batch creation of virtual machines, return the deployment information of the virtual machine, including the running physical server, used PCI card and network connection information, for further management and maintenance.

[0019] As a preferred technical solution of the present application: the scheduling algorithm includes the following formula to schedule based on PCI card bandwidth:

[0020]

[0021] Where bwpci represents the bandwidth of the current host node PCI card, and bwmin is the minimum bandwidth required by the virtual machine. The formula calculates the score of the PCI card attribute, and adds the host nodes that meet the conditions to the schedulable list.

[0022] As a preferred technical solution of the present application: the scheduling method creates a new filter class 'DPciPassthroughFilter' by extending the Nova scheduler, which filters the physical host nodes according to the model and bandwidth information of the PCI card.

[0023] As a preferred technical scheme of the present application: the filter type implementation is to inherit the 'nova.scheduler.filters.BaseHostFilter' class, and implement the 'host_passes' method therein, and schedule the nodes meeting the requirements by obtaining the PCI card model and bandwidth information of each host node.

[0024] As a preferred technical scheme of the present application: the batch scheduling algorithm includes the following formula to schedule based on network topology:

[0025]

[0026] Wherein, di is the network distance between each host node, and dideal is the ideal network distance, and the network topology matching degree of multiple host nodes is calculated through the formula to ensure that the scheduling result meets the network connection requirements.

[0027] As a preferred technical scheme of the present application: the batch scheduling algorithm extends the Nova scheduler and creates a batch filter class 'BaseBatchHostFilter', which filters the host node combination meeting the network topology requirements through the interface 'hosts_passes'.

[0028] As a preferred technical scheme of the present application: the virtual machine resource requirement list includes the following information: CPU resource requirement, memory resource requirement, physical properties of PCI card, network connection relationship between virtual machines, etc.

[0029] As a preferred technical scheme of the present application: the scheduling algorithm combines the native scheduling algorithm of the cloud platform with the PCI card attribute scheduling and network topology batch scheduling function to calculate the best deployment position of the virtual machine in the physical server.

[0030] As a preferred technical scheme of the present application: when performing scheduling, according to the formula:

[0031] f(node s election)=α·f(pci s core)+β·ftopology m atch)

[0032] Wherein, α and β are the weight coefficients of PCl attribute and network topology matching degree, respectively, and the comprehensive score of each host node is calculated, and the node with the highest score is finally selected for virtual machine scheduling.

[0033] As a preferred technical scheme of the present application: the scheduling algorithm dynamically updates the resource state of the physical server in real time during the scheduling process, so as to ensure that the scheduling result can adapt to the real-time network and hardware resource changes.

[0034] Compared with the prior art, the beneficial effects of the present application are:

[0035] 1、The present application significantly reduces the delay problem in data transmission and improves network performance by introducing PCI pass-through technology and accurately filtering according to PCI network card attributes (such as bandwidth, model) during virtual machine scheduling. By directly accessing physical network cards, the performance loss caused by the virtualization layer is avoided, ensuring that virtual machines in the railway communication system can efficiently process and transmit large data, and realizing the real-time data processing requirements in the train operation process.

[0036] 2、Although the PCI pass-through technology does not support traditional virtual network security group functions, the present application ensures the isolation and security between virtual machines through the configuration of physical switches, combined with network topology and scheduling algorithms. When creating virtual machines in batches, the scheduling algorithm realizes accurate deployment and network isolation of virtual machines based on the network topology of the physical layer, ensuring the security requirements of the railway communication system.

[0037] 3、The present application not only schedules based on PCI attributes, but also extends the batch scheduling function, enabling it to schedule virtual machines in batches according to the specific network topology of the railway communication system. Through the network topology matching algorithm, it ensures that the deployment of virtual machines meets the specific network connection requirements, improving the flexibility and precision of scheduling, thereby meeting the high standards of real-time performance and reliability in complex network environments of the railway communication system. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 A method flowchart of a cloud computing virtual machine scheduling method suitable for railway communication cloud. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0040] The specific embodiments of the present application will be described in detail below in conjunction with the accompanying drawings and multiple embodiments.

[0041] Embodiment 1: Virtual machine scheduling based on PCI network card attributes

[0042] Reference Figure 1In the deployment of the railway communication cloud, a virtual machine needs to process high-bandwidth video stream data, with a bandwidth requirement of at least 10 Gbps, and the virtual machine needs low-latency network transmission. To meet this requirement, the system administrator defines a resource requirement list for the virtual machine, specifically describing the CPU, memory, and network card bandwidth requirements (10 Gbps or higher) of the virtual machine.

[0043] First, the cloud platform collects resource information of all physical servers, including the model and bandwidth of the PCI network card. According to the PCI card attribute scheduling algorithm in the claims, the following formula is used to calculate the PCI bandwidth:

[0044]

[0045] where bw pci is the bandwidth of the PCI network card on the physical server, and bw min is 10 Gbps. This formula scores the PCI card bandwidth of all physical servers, and only selects physical server nodes that meet the bandwidth requirement of 10 Gbps or higher.

[0046] Through the filter 'DPciPassthroughFilter', the physical server A with a bandwidth of 25 Gbps is finally selected, and the virtual machine is scheduled to this server, ensuring that the virtual machine can obtain sufficient network bandwidth to meet the efficient transmission of video streams.

[0047] Embodiment 2: Batch scheduling based on network topology

[0048] Referring to Figure 1 , a railway communication system needs to deploy a group of virtual machines for real-time data synchronization and scheduling control between stations.

[0049] The system requires a star-shaped network topology, and the connection between virtual machines is through the central server (master node) for communication, and the network transmission delay must be controlled within 10 milliseconds.

[0050] The administrator first defines the network topology structure of the system through the cloud platform, specifying the number of virtual machines, the network type (Ethernet) of each virtual machine, the device type (such as master node device, edge device, etc.), and the connection relationship.

[0051] Next, the cloud platform collects network connection information of each physical server, including the physical distance and network delay between servers. According to the network topology requirements of the virtual machine, the scheduling algorithm matches the network topology through the following formula:

[0052]

[0053] where di d is the network distance between servers ideal d is the ideal network latency between virtual machines (10 ms). Through the batch scheduling algorithm BaseBatchHostFilter, the server combination that meets the virtual machine network latency requirements and conforms to the star topology is screened out, and the virtual machines are batch scheduled to these servers, ensuring that each virtual machine can be correctly connected to the master node device and meet the real-time data synchronization requirements.

[0054] Example 3: Real-time adjustment of physical server resource state scheduling

[0055] Referring to Figure 1 In the railway communication cloud, as the real-time data transmission demand of train operation increases, the virtual machine load on a certain server group gradually rises, causing the resource usage to approach saturation.

[0056] At this time, the cloud platform needs to automatically adjust the deployment location of the virtual machine to avoid performance bottlenecks and maintain the efficient operation of the system.

[0057] The cloud platform monitors the resource state of the physical server and finds that the CPU and memory usage of some servers exceeds 80%. According to the scheduling algorithm in the claim, the system will dynamically update the resource information of each physical server in real time and recalculate the optimal deployment location of the virtual machine. The following formula is used for comprehensive scoring:

[0058] f(node s election)=α·f(pci s core)+β·f(topology m atch)

[0059] Where α and β are weight coefficients, respectively reflecting the importance of PCI attributes and network topology.

[0060] The formula takes into account the PCI network card bandwidth, network topology matching degree, and current server load state. Finally, the system selects another physical server B, which has low resource utilization and meets the network topology requirements.

[0061] The virtual machine is automatically migrated to server B, maintaining the network performance of the railway communication system.

[0062] Example 4: Large-scale virtual machine deployment in railway communication cloud

[0063] Referring to Figure 1 A railway communication system needs to deploy a large-scale virtual machine group in the newly opened line to handle big data analysis tasks generated during train operation. The system requires that all virtual machines must meet the bus topology and the network bandwidth requirement is at least 20 Gbps.

[0064] The administrator first defined the resource requirements list and bus topology of the virtual machines, and the cloud platform began collecting resource information from the physical servers. The scheduling algorithm first filtered out physical server nodes with bandwidth of 20Gbps or higher using the 'DPciPassthroughFilter', and then, based on the bus network topology requirements, used a batch scheduling algorithm to select combinations of server nodes that could meet the requirements of that network topology.

[0065] During the scheduling process, the cloud platform combines PCI network card attributes and network topology, and uses the following formula for comprehensive scoring.

[0066] f(nod selection)=α·f(pciscore)+β·f(topology mach)

[0067] Ultimately, the system deployed virtual machine clusters on multiple physical servers in batches and automatically configured network connections to ensure that all virtual machines could be correctly connected in a bus topology and meet the needs of big data analysis tasks.

[0068] Example 5: Fault Tolerance Mechanism in Railway Communication Systems

[0069] See Figure 1 The high security and reliability requirements of railway communication systems necessitate that virtual machines automatically migrate to other servers to maintain normal system operation when some physical servers fail. To this end, the railway communication cloud has deployed a scheduling algorithm based on this invention, supporting automatic fault tolerance.

[0070] After server C fails, the cloud platform automatically detects that the server is unavailable. The system monitors the server's resource status and updates the resource information of other servers in real time. The scheduling algorithm is executed again, and based on the resource status of the remaining available servers, PCI network card bandwidth, and network topology matching degree, a new target server is selected, and the affected virtual machines are automatically migrated to another server D that meets the conditions.

[0071] Through this fault-tolerance mechanism, the railway communication system can maintain service continuity even when the server fails, and ensure that the network connectivity and performance of the virtual machine meet the preset requirements.

[0072] The content not described in detail in the specification belongs to the prior art known to those skilled in the art, although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A cloud computing virtual machine scheduling method suitable for railway communication cloud, characterized in that: The method comprises the following steps: Step 1: Collecting hardware resource information of each physical server in the cloud platform, the hardware resource information including CPU, memory, model of PCI network card, bandwidth, physical location and network connection topology basic information; Step 2: Inputting and defining network topology structure of the railway communication system by the user, the topology structure including network type, topology form, device type and network connection demand thereof, the defined network topology determining network connection and resource allocation among virtual machines; Step 3: Generating resource demand list of the virtual machine according to the network topology structure defined by the user and the resource information of the physical server, the resource demand list including required CPU, memory, PCI network card bandwidth, network interface type and connection relationship among virtual machines; Step 4: Screening the physical server meeting the demand according to the resource demand list of the virtual machine by using a scheduling algorithm, the algorithm first screening according to the PCI network card attribute; Step 5: After screening the physical server meeting the PCI network card attribute requirement, further screening the physical server according to the network topology demand of the virtual machine by using a batch scheduling algorithm; the batch scheduling algorithm including the following formula to schedule based on the network topology: wherein, is the network distance between each physical server, is the ideal network distance, the network topology matching degree of the plurality of physical servers is calculated by the formula; Step 6: Batch creating the virtual machine on the physical server selected through the above screening steps, and establishing correct network connection between each virtual machine and its corresponding physical network card and network connection topology, so as to ensure that the virtual machine can normally run and meet the requirements of the railway communication system on network performance and isolation; Step 7: Returning the deployment information of the virtual machine including the running physical server, used PCI network card and network connection information after completing the batch creation of the virtual machine. 2.The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 1, wherein: The scheduling algorithm of step 4 includes the following formula to schedule based on the PCI network card bandwidth: Wherein, represents the bandwidth of the PCl network card of the current physical server, is the minimum bandwidth required by the virtual machine, the score of the PCI network card attribute is calculated by the formula, and the physical server meeting the conditions is added to the schedulable list. 3.The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 2, wherein: The scheduling method expands the Nova scheduler to create a new filter class 'DPciPassthroughFilter', which screens the physical server according to the model and bandwidth information of the PCI network card.

4. The cloud computing virtual machine scheduling method of claim 3, wherein: The implementation mode of the filter class is to inherit the 'nova.scheduler.filters.BaseHostFilter' class and implement the 'host-passes' method therein, which schedules the physical server meeting the requirement by obtaining the model and bandwidth information of the PCI network card of each physical server.

5. The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 1, characterized in that: The batch scheduling algorithm expands the Nova scheduler to create a batch filter class 'BaseBatchHostFilter', which batch screens the physical server combination meeting the network topology requirement through the interface 'hosts-passes'.

6. The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 1, characterized in that: The scheduling algorithm combines the native scheduling algorithm of the cloud platform with the PCI network card attribute scheduling and network topology batch scheduling function to calculate the best deployment position of the virtual machine in the physical server.

7. The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 1, characterized in that: In the execution of scheduling, according to the formula: wherein, and are respectively the weight coefficients of PCl attribute and network topology matching degree, the comprehensive score of each physical server is calculated, and the physical server with the highest score is finally selected for virtual machine scheduling. 8.The cloud computing virtual machine scheduling method suitable for railway communication cloud according to claim 7, wherein: The scheduling algorithm dynamically updates the resource state of the physical server in real time during the scheduling process, so as to ensure that the scheduling result can adapt to the real-time network and hardware resource changes.

Images