CF Server-Less: CloudFoundry as a Platform for Server-less Computing
summary
Implements a CF Server-Less layer that turns CF apps on and off to emulate functions, deploys functions in Python with 512 MB in US East, and runs 100-request experiments sequentially and in parallel across OpenWhisk, Lambda, Azure and Iron.io. Observations: echo high‑throughput averages ~180–200 ms but cold‑start spikes occur; low‑throughput shows providers either kill containers (long cold start) or freeze with cgroups (hundreds of ms resume); memory test ~1.3 s average; CPU test ~2.5 s on CF sequential, parallel cases rise to ~8–9 s; Azure showed severe failures (only ~15% success) under memory load.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:20Surveillance Computing
- 2:27Why not use Cloud Foundry
- 3:50What are our goals
- 5:29Approach
- 6:00Environments
- 6:46Surveillance
- 11:34Demo
- 13:00Results
- 13:59Parallel
- 17:31CPU Intensive Performance
- 17:52Open Risk Performance
related talks
25mCF Serverless: CloudFoundry as a Platform for Serverless Computing
23mServerless in the Wild: Characterizing and Optimizing the Serverless Workload
18mServerless (aka FaaS) Computing With Swift
27mServerless Performance on a Budget
37mGo FaaSter: Cold Start Optimization in a Serverless Platform
31mPut Your Money Where Your CF Is
37mGo FaaSter: Cold Start Optimization in a Serverless Platform
37mState of Serverless