HTTP request proxying by a web accelerator
summary
Describes troubleshooting proxying HTTP requests through a web accelerator (HTTP accelerator) placed in front of a web server when user requests still experience high latency or the backend appears underutilized. Covers cases where introducing the accelerator causes rare, hard-to-reproduce failures in the web application that users notice but operators may not, and discusses diagnosing and resolving such proxy-accelerator interaction problems.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:29HTTP proxying problems
- 3:58Keep-Alive connections
- 6:05Optimal connection count
- 8:34Connection drops
- 9:39Server health checks
- 11:27Request pipelining
- 12:56Pipelining security
- 16:36Unreachable requests handling
- 19:06Queues and load balancing
- 24:12Failed request handling
- 25:13Double buffering issue
- 28:27Cryptography problems
- 31:17Buffering vs streaming
- 32:47Data copy problem
- 35:19Kernel copy optimization
- 37:13Fast kernel transport
- 38:25Tempesta FW architecture
- 40:34HTTP/2 features
- 41:58HTTP/2 in accelerator
- 42:42Current status and plans
related talks
42mHow a Web Accelerator Accelerates Your Site
43mWeb Acceleration Mechanics
34mHow HTTP already solved all your performance problems 10 years ago
39mHTTP Caching for REST APIs
51mBetter by Proxy
28mScaling Applications with AWS Global Accelerator and Amazon CloudFront
37mRunning JS on the Server at Scale — What Can Go Wrong?
32mAll the Proxies!