Lightning Talk: Performance Testing of FoundationDB using TPC-C

on YouTube ↗06m

" summary

Snowflake uses a TPC-C workload and periodic Jenkins runs to surface regressions and measure feature changes against a 5-node AWS cluster with triple replication, three resolvers, three proxies and five transaction logs. Tests reveal linear scaling at 32 and 64 storage servers but a surprising throughput drop at 128 storage servers caused by increased real latency, longer transactions and higher optimistic-concurrency conflicts. Optimizations—pre-serialization, throttling changes, pushing metrics to proxies and a cluster-controller class—raise status handling by about 150% more status requests per second.

AI-written summary · about

" chapters (6)

1. 0:00Introduction
2. 0:49How we use TPCC
3. 1:29Test setup
4. 3:21Cluster controller CPU
5. 4:07Optimizations
6. 5:23Future Plans

" related talks

• 18mPerformance Profiling in FoundationDBDiego Didona · FoundationDB Summit 2019
• 06mLightning Talk: Performance Engineering and Testing of Apps and Services Deployed on Cloud FoundrySiva Balan · Cloud Foundry Summit Silicon Valley 2017
• 34mFoundationDB, the Black KnightPeter Boros · Open Source Summit North America 2025
• 06mLightning talk: Testing Distributed DatabasesMark Nadal · Nordic.js 2017
• 56mLightning TalksMark Wong, Chris Ellis, Jelte Fennema-Nio, Claire Giordano, Neeta Goel, Hannu Krosing, Christoph Berg, Di Qi, Enes Cakir, Raoul De Guchteneere, Marat Bogatyrev, Eren Basak · PostgreSQL Conference Europe 2024
• 40mPerformance Testing and Big DataRoland van Leusden · TestCon Vilnius 2016
• 16mFoundationDB and Node.js: Year in ProductionSteve Korshakov · FoundationDB Summit 2019
• 40mHow DTCC achieves data resiliency with Snowflake and AWS (PRT272)Arjun Puri, Sudha Gullapalli · AWS re:Invent 2022