gc-latency

README

Writing GC Friendly Code

Showing how you to use GC friendly data.

Use case: User store that returns user (string) from user ID (int).

• users_map uses a map[int]string
• users_slice use a []string where the location on the slice is the ID
• users_str builds one big string with all user names and also have indices []int which hold the end index for each user

To bench a specific implementation, say "map", run:

make map
# In another terminal
make stress

FIXME: I'm not sure that speedup is due to GC. If I run the benchmark in each implementation I see on my machine:

• map : BenchmarkByID-12 146152878 8.117 ns/op 0 B/op 0 allocs/op
• slice: BenchmarkByID-12 871609362 1.379 ns/op 0 B/op 0 allocs/op
• str : BenchmarkByID-12 524216815 2.309 ns/op 0 B/op 0 allocs/op

And for make stress I see:

• map : Requests/sec: 139456.5474
• slice: Requests/sec: 131119.3114
• str : Requests/sec: 141903.6378

---

Trace

• Add import to net/http/pprof
• go run ./httpd
• hey -z 10s http://localhost:8080/users/353
• curl -o trace.out http://localhost:8080/debug/pprof/trace?seconds=5
• go tool trace -http=:8888 trace.out
  • trace.png
• 1,111,313ns -> 1.1ms