Profiling / Optimising

[zanuka]

The Go app should leverage the built-in profiler that supports CPU, memory, goroutine and block (contention) profiling.

easily enabled:

import (
    "net/http"
    _ "net/http/pprof"
)

advantages:

• low overhead
• can be used in a production environment
• can be used with live traffic without any noticeable performance penalties

dev tasks (issues):

• report the number of incoming requests
• report how long every request takes.
• report request details - (URL, IP address and user-agent)
• send all stats to metric aggregator (e.g. StatsD) In addition, the service needs to
• measure how many requests per second the API can handle (Apache Benchmark tool)
• profile the CPU to determine which functions spend most time
• run heap profiler for in-use memory and number of allocated objects
• block profiling to show function calls that lead to blocking on sync primitives like mutexes and channels