SLATE: Globally Optimizing Request Routing

on YouTube ↗12m

" summary

SLATE builds a global controller plus per-cluster controllers and a proxy-wasm HTTP filter that reports per-service RPS and trace latencies, then runs a linear programming optimizer over those profiles to partition traffic across regions. The prototype reduces cross-cluster egress and average latency by pushing routing percentages (example: it reroutes about 25% of product-page requests away from overloaded clusters), and it translates optimizer output into virtual service rules that the cluster controllers enforce.

AI-written summary · about

" chapters (7)

1. 0:00SLATE introduction
2. 2:07Baselines and optimization
3. 3:11System architecture
4. 4:57WebAssembly plugin
5. 6:00Latency optimization
6. 8:56Prototype results
7. 10:00Future directions

" related talks

• 18mSLATE: Service Layer Traffic EngineeringAditya S Prerepa, Gangmuk Lim, Brighten Godfrey, Radhika Mittal · NSDI '26
• 21mServiceRouter: Hyperscale Service MeshHarshit Saokar, Margot Leibold · Systems @Scale Summer 2023
• 21mHandling Traffic Spikes Through Automated Prioritized Load SheddingBenjamin Fedorka · gRPConf 2025
• 16mTowards Optimal Rack-scale μs-level CPU Scheduling through In-Network Workload...Kai Chen, Hao Wang, Han Tian, Xudong Liao, Junxue Zhang, Guyue Liu, Xinchen Wan, Chaoliang Zeng, Mengyu Ma · USENIX ATC '25
• 17mServiceRouter: Hyperscale and Minimal Cost Service MeshHitesh Khandelwal, Harshit Saokar, Margot Leibold, Soteris Demetriou, Chunqiang Tang, Dimitrios Skarlatos, Nick Magerko, Max Kontorovich, Josh Kirstein · OSDI 2023
• 20mBetter Resilience Through Self TuningWilliam Thurston · Systems @Scale Summer 2021
• 21mR2P2: Making RPCs first-class datacenter citizensAdrien Ghosn, Marios Kogias, Edouard Bugnion, George Prekas, Jonas Fietz · USENIX ATC '19
• 17mLibra: Flexible Request Partitioning and Scheduling for Serving Unbalanced and Dynamic...Cheng Li, JIALIN LI, Yongji Wu, Chaoyi Ruan, Yinhe Chen, Dongqi Tian, Yandong Shi · NSDI '26