Down-to-Earth Cloud Scaling
summary
Cloud projects test scaling late and assume the cloud fixes architecture; load tests expose that App Service plans can scale out but additional instances never start properly, and a parallel indexing rewrite exhausts datastore socket limits so a design that is too fast fails. Real fixes: automated load scripts (developers ran one over a holiday), realistic datasets (Reuters images extrapolated from ~1k to millions), metrics in a consolidated Grafana dashboard, and tuning Cosmos DB throughput. Continuous testing and iterative infra changes are mandatory.
AI-written summary · about
chapters (25)
- 0:00Intro
- 2:25React to change
- 4:56Do you know your system will scale
- 7:04When to test
- 10:33Why testing
- 14:04Close to production specs
- 15:00Load testing
- 23:07Designing load tests
- 24:38Frontloading features
- 26:23infrastructure
- 27:33design guidelines
- 29:19Kill your Darlings
- 30:31Linear growth
- 31:31Design questions
- 33:21Scaling up
- 34:24Scaling out
- 34:56Scaling throughput
- 35:54Scaling unrelated tasks
- 38:27When to scale
- 39:27Monitoring
- 40:39Dashboard
- 41:58CPU Memory Usage
- 42:35App Insights
- 42:48Change your monitoring
- 45:10Recap
related talks
01h 08mPerformance testing in the cloud
57mLast- und Performancetests in der Cloud?!
37mA startup's infrastructure - how to make scaling a sales, not a tech issue
32mScaling Done Right
40mScaling Startup for Success: Building modern Architectures and Developer Acceleration
27mTesting Before You Scale & Making Friends While You Do It
12mExponential Load Testing...
31mAutoscaling Apps without Supersizing Costs