FaaSNet: Scalable and Fast Provisioning of Custom Serverless Container Runtimes...

on YouTube ↗12m

" summary

FaaSNet implements a lightweight, hierarchical function tree plus an I/O‑efficient on‑demand fetching protocol inside Alibaba Cloud Function Compute to cut image pull bottlenecks. It provisions 2,500 function containers on 1,000 VMs in 8.3 seconds, scales 13.4× and 16.3× faster than Alibaba’s prior FaaS platform and Kraken respectively, and sustains bursty workloads using 75.2% less time than an optimized baseline; prototype and integration with Alibaba Cloud are implemented and released.

AI-written summary · about

" chapters (6)

1. 0:00Introduction to serverless
2. 1:03Custom container runtimes
3. 2:30Workload and latency analysis
4. 4:23Fastnet architecture design
5. 6:56Evaluation and performance
6. 10:51Conclusion and resources

" related talks

• 51mFunction as a Service in private cloudSergey Rybalkin · HighLoad++ Moscow 2018
• 35mAchieving Faas + Dapr + K8S technology integration, and build a new Alibaba Cloud Se...Zhao Qingjie · KubeCon + CloudNativeCon + Open Source Summit China 2021
• 31mGoing FaaSter: Function as a Service at NetflixYunong Xiao · Node+JS Interactive 2018
• 34mFaaS and Furious - 0 to Serverless in 60 Seconds, AnywhereAlex Ellis · KubeCon + CloudNativeCon North America 2017
• 37mGo FaaSter: Cold Start Optimization in a Serverless PlatformScott Zhou, Yanbo Li · KubeCon + CloudNativeCon China 2019
• 37mGo FaaSter: Cold Start Optimization in a Serverless PlatformScott Zhou, Yanbo Li · KubeCon + CloudNativeCon + Open Source Summit China 2019
• 23mFaasm: Lightweight Isolation for Efficient Stateful Serverless ComputingPeter Pietzuch, Simon Shillaker · USENIX ATC '20
• 26mGoing Faaster: Function as a Service at NetflixYunong Xiao · NodeConf EU 2018