Statistics Aggregator for Network Virtualization Query Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network virtualization platforms face inefficiencies in retrieving interface statistics, particularly due to slow mechanisms for collecting data, which hinder the ability to provide a bulk API for retrieving statistics across multiple hypervisors, leading to impractical and slow operations.
Innovation Solution
A statistics collection framework that periodically aggregates interface statistics, such as flow counter information, using a statistics aggregator, which tags additional identification information to manage discontinuities and acts as a cumulative cache, allowing for faster query responses by eliminating the need to fetch data across multiple hypervisors simultaneously.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistics are retrieved on demand from multiple hypervisors, then the system can provide accurate real-time data, but the query speed becomes slow and impractical
Solution Approach 1:
The system performs preliminary actions by periodically collecting and pre-storing interface statistics in a cumulative cache within the statistics aggregator before queries are made. This allows the aggregator to have statistics ready in advance, eliminating the need to fetch data across multiple hypervisors at query time and enabling fast bulk retrieval operations.
2Speed
If statistics are collected periodically and cached, then query speed improves, but the data may not be fully real-time accurate
Solution Approach 1:
The statistics aggregator acts as an intermediary between the hypervisors and the querying system. It maintains a cumulative cache that periodically receives and aggregates interface statistics from multiple hypervisors, then serves this pre-aggregated data to queries. This intermediary approach balances real-time accuracy requirements with query performance by providing near-real-time data without requiring simultaneous access to all hypervisors.
3Quantity of substance
If bulk queries retrieve all interface statistics across multiple hypervisors, then complete data is obtained, but the operation becomes slow and impractical
Solution Approach 1:
The statistics aggregator performs preliminary aggregation of interface statistics from multiple hypervisors into a cumulative cache before bulk queries are executed. When bulk queries request all interface statistics, the aggregator serves this pre-aggregated data locally, eliminating the time-consuming operation of fetching and aggregating data from multiple hypervisors at query time while still providing complete data coverage.
Data Source
AI summary
Some embodiments provide a statistics collection framework that is used to aggregate statistic for interfaces such as logical ports and logical port pairs. Flows that are related with these interfaces are tagged with the identifier of the logical entities for which statistics are being collected. The interface statistics is periodically sent in the background to a statistics aggregator. The read queries for the interface statistics are directed to the statistics aggregator. The statistics aggregator, therefore, acts as a cumulative cache for the interface statistics.