You can download the slides from this presentation here.
In a recent technical presentation, Tim Carlin, QA Cafe's Senior Developer, offered a comprehensive analysis of CDRouter's Differential Services Code Point (DSCP) testing methods. He explained that DSCP testing evaluates Quality of Service (QoS) policies by measuring and categorizing data flow.
Carlin provided a detailed explanation of test setups, stressing the necessity of isolating testing environments to avoid extraneous variables. He demonstrated how video and voice data flows utilized approximately 50% and 25% of the overall measured throughput, respectively, with the remainder divided between other 'best effort' flows. He clarified that the resulting 54% data loss was anticipated due to the transmission being at 100% of measured capacity but only allowing 50% throughput.
Using CDRouter, he performed live DSCP tests to highlight certain points. Results from these tests revealed the critical nature of queuing in managing data flow and indicated some challenges in downloading test cases. He cautioned against mixing upload and download test cases and provided an overview of CDRouter flags to assist in troubleshooting.
Carlin also covered the issues surrounding performance testing, noting it to be a moving target due to its inherent variability and the constant drive for speed improvement. He emphasized that CDRouter aids in creating repeatable, consistent results and helps define meaningful expectations, thresholds, and targets.
Furthermore, he cautioned that queuing can be CPU intensive and that test case changes could affect results. However, he confirmed that CDRouter's traffic flows and thresholds assist in maintaining consistent results. He also suggested combining this approach with stability testing to examine device performance over time and under additional processing requirements.
Carlin concluded by referring to additional resources that offer a deeper understanding of DSCP testing, including the RFCs on queueing and per-hop behavior, OpenWRT's queuing model, and QA Cafe's own guides on performance and DSCP testing.
