"Performance Testing: ..."
summary
Real user monitoring across 500 million interactions on ~200 sites finds that shaving 2 seconds off response time cuts page abandonment ~8% (an 8‑second improvement cuts abandonment ~38%), and typical transactions call 8.8 hosts per transaction. Cloud dependency grows—about 20% of business transactions touch EC2-hosted objects—and 17 months of identical-load tests show large inter‑ and intra‑cloud variability and lower than acceptable availability. A community tool cloud sleuth clouds.net collects signals; an S3 sharding experiment yields a 50% performance gain at no extra data cost.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:28Cloud
- 2:53Question
- 5:39Cloud Characteristics
- 6:48Availability
- 7:23Cloud is opaque
- 8:13Cloud Sleuth
- 8:55Cloud Optimization
- 11:41Sharding
- 12:50Aggregate Information
- 13:41Summary
related talks
57mLast- und Performancetests in der Cloud?!
49mEvery Millisecond Counts Performance & UX
09mPerformance is the first UX challenge to consider
01h 01mWeb Performance Deep Dive
21mPerformance Consistency in the Cloud
56mBetting on Performance: A note on Hypothesis-driven Performance Testing
31mEvery Millisecond Counts: Performance & UX
57mHigh Performance Mobile