Distributed Network Analytics Maplets for Adaptive Data Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional computer network analytics require users to predefine data generation and aggregation queries, leading to inefficient resource usage as all measurements are collected simultaneously, even if not all are needed, wasting network resources.
Innovation Solution
A distributed system decomposes network analytics tasks into subtasks called maplets, which are deployed across network nodes, allowing for adaptive refinement of analytics tasks and data source configuration based on collected data, enabling dynamic adjustment of data generation and collection granularity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all measurements are collected simultaneously in traditional network analytics, then complete data coverage is achieved, but network resources are wasted due to collecting unnecessary data
Solution Approach 1:
The patent segments the network analytics system into distributed agents deployed at network nodes, where each agent independently collects and processes measurements. This segmentation enables selective data collection based on local conditions, avoiding the waste of collecting all measurements across the entire network simultaneously.
Solution Approach 2:
The system dynamically adjusts data collection based on detected events. When an event is detected by any agent, the system adapts by collecting additional granular data from relevant sources. This dynamic approach ensures complete data coverage only when necessary, optimizing network resource usage.
2Ease of operation
If users predefine all aggregation queries beforehand, then query execution is simplified, but adaptability to emerging patterns is reduced
Solution Approach 1:
The system implements feedback mechanisms where agents continuously monitor network conditions and report findings to the controller. Based on this feedback, the controller dynamically generates and refines aggregation queries, enabling the system to adapt to emerging patterns while maintaining ease of operation through automated query management.
Solution Approach 2:
The system performs preliminary data collection and event detection at distributed agents before centralized query execution. This preliminary action enables the system to pre-process data locally and only trigger detailed aggregation queries when events are detected, combining simplified execution with adaptive response to emerging patterns.
3Reliability
If data sources are activated continuously, then data availability is maximized, but system resource consumption increases
Solution Approach 1:
The patent implements periodic event-driven activation of data sources rather than continuous activation. Data sources are activated periodically based on event detection, ensuring data availability when needed while minimizing resource consumption during normal operation. This periodic action maintains reliability by activating sources systematically when events occur.
Data Source
AI summary
A network analytics controller in a computer network receives specification of a network analysis task. The controller creates maplets. Each maplet corresponds to an agent executing on a network node in the computer network. Each maplet specifies the network data collection configuration, collected data processing, and processed data reporting to be performed to support the network analysis task at the node executing the corresponding agent. The controller communicates the maplet corresponding to the agent over the computer network to each agent. Each agent receiving a maplet configures its corresponding network node for network data collection, collected data processing, and processed data reporting in accordance with the received maplet. Each agent receiving a maplet collects and processes network data, and reports the processed data to the controller, in accordance with the configuration. The controller aggregates, as the result of the network analysis task, the data reported from each agent.