Network SLOs: Knowing When The Network Is The Barrier To Application Quality Of Experience
summary
Meta defines concrete Network SLOs by measuring passive transfer times of synthetic blocks (not speedtests) and choosing the best predictor via AUC to map transfer-time thresholds to bad-session rates. Measurements use 50KB and 200KB block sizes, cluster clients into 15-minute network aggregates, and declare an aggregate SLO-compliant if a set percentage of samples meet the SLI. This system runs on trillions of measurements per day, surfaces edge-capacity gaps, and successfully triages network vs product QoE regressions.
AI-written summary · about
chapters (13)
- 0:00Network SLOs and QoE
- 1:07Network requirements for online products
- 3:36Defining Network SLOs
- 4:39Measuring quality of experience in video applications
- 5:42App startup performance
- 6:45Characterizing network conditions
- 8:52Synthetic blocks for network measurement
- 12:22Network SLOs from QoS
- 15:55Network aggregates for compliance
- 18:23SLOs to improve network readiness
- 20:09Edge capacity from SLOs
- 21:57Recap and applications
- 22:39Q&A
related talks
25mResQ: Enabling SLOs in Network Function Virtualization
37mA Bartender's Guide to Network Monitoring
17mHow to diagnose nanosecond network latencies in rich end-host stacks
46mSelf-Programming Networks: Network Clock Synchronization, Event Playback and Real-time Action
32mLive Q+A 2
45mReducing Network Latency: Innovations for a Faster Internet
27mSIMON: A Simple and Scalable Method for Sensing, Inference and Measurement in...
46mLessons learned from monitoring a large production network