Network Switch Performance Monitoring via Triggered Buffer Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current performance monitoring systems in high-performance computers face inefficiencies due to high overhead and imprecision in reading switch performance counters, leading to distorted data and reduced statistical quality, especially when measuring multiple switches simultaneously.
Innovation Solution
A communication system that uses request messages with read triggers, counter selectors, and completion triggers to collect and store counter data in a local buffer, allowing for periodic sampling and reduced message transmission, thereby improving data quality and reducing network overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If performance counters are read frequently to improve measurement precision, then data accuracy is improved, but network overhead increases and measurement distortion increases
Solution Approach 1:
The switch stores counter samples in a local buffer before they are requested by the core interface. This preliminary action allows the core interface to read multiple counter values without sending multiple read requests to the switch, reducing network overhead while maintaining measurement precision through the use of triggers that coordinate reading operations.
2Productivity
If multiple switches are measured simultaneously to improve system monitoring capability, then productivity is improved, but measurement precision deteriorates due to distorted data
Solution Approach 1:
The system uses completion triggers that provide feedback to the core interface about when counter reads have been performed. This feedback mechanism allows the core interface to coordinate read operations across multiple switches, ensuring that counter readings are captured accurately without distortion from simultaneous measurement traffic, while still enabling monitoring of multiple switches.
Solution Approach 2:
Each switch performs preliminary counter sampling and stores results in a local buffer before the core interface reads them. This preliminary action decouples the measurement process from the network transmission process, allowing multiple switches to be monitored simultaneously without the measurement traffic itself causing distortion in the counter data.
3Measurement precision
If counter reading operations are performed frequently to improve measurement quality, then measurement precision is improved, but loss of time increases due to repeated measurements
Solution Approach 1:
The switch performs counter sampling in advance and stores the results in a local buffer. When the core interface requests data, it can read multiple counter values from the buffer without waiting for repeated measurements. This preliminary sampling action improves measurement precision while reducing the time required for subsequent readings.
Solution Approach 2:
The counter readings are performed continuously and stored in a buffer, creating a continuous stream of measurement data. This continuous action ensures that measurement precision is maintained through consistent sampling, while the buffered data can be read efficiently without repeated measurement operations, reducing time loss.
Data Source
AI summary
Particular embodiments described herein provide for a network element that can be configured to receive a request message, wherein the request message includes a read trigger, an indicator selector, and a completion trigger, determine an indicator that relates to the indicator selector, and perform an action when the read trigger is activated.