Scaling IO-bound application in the cloud
summary
Claim: storage, not CPU or RAM, often becomes the scaling bottleneck. Demonstrates with fio random-write 60k IOPS on a laptop and an UpCloud cheap VM, while an older HDD backend hits ~1k IOPS, so raw disk speed radically changes real-world app latency. Shows quick provisioning via init scripts and API (server ready in ~1 minute) and discusses practical fixes: sharding/partitioning, separate write logging, caching; a working demo of deployment and benchmarks validates the performance differences.
AI-written summary · about
chapters (4)
- 0:00Start
- 6:24Scaling Apps in the Cloud
- 16:00Tips for sharing the 10 load
- 16:19Divide et Impera - Partitioning
related talks
47mExploring Linux Memory Usage and IO Performance in Cloud Native Applications
17mλ-IO: A Unified IO Stack for Computational Storage
40mBenchmarking: Repeatable & Comparable
30mFast, Faster, Fastest: Object Storage, Cloud Block Storage, and SSD
04mLightning Talk - Benchmarking Persistent Data Volumes
40mAchieving the Ultimate Performance with KVM
57mLast- und Performancetests in der Cloud?!
30mAbove the Clouds: Investigating Systems from 32,000 ft.