Switch to uint64 hash for identifier

[dashpole]

goos: linux
goarch: amd64
pkg: github.com/GoogleCloudPlatform/opentelemetry-operations-go/exporter/collector/internal/datapointstorage
cpu: AMD EPYC 7B12
             │ bench2.txt  │             bench3.txt             │
             │   sec/op    │   sec/op     vs base               │
Identifier-2   4.805µ ± 4%   2.539µ ± 0%  -47.16% (p=0.002 n=6)

             │ bench2.txt  │            bench3.txt             │
             │    B/op     │    B/op     vs base               │
Identifier-2   2208.0 ± 0%   632.0 ± 0%  -71.38% (p=0.002 n=6)

             │ bench2.txt │            bench3.txt            │
             │ allocs/op  │ allocs/op   vs base              │
Identifier-2   47.00 ± 0%   44.00 ± 0%  -6.38% (p=0.002 n=6)

I'm still working on testing edge cases to make sure identifiers don't collide.

[codecov]

Codecov Report

Attention: Patch coverage is 91.11111% with 4 lines in your changes missing coverage. Please review.

Project coverage is 62.97%. Comparing base (4caace7) to head (7de96dd).
Report is 38 commits behind head on main.

Files with missing lines	Patch %	Lines
...ctor/internal/normalization/disabled_normalizer.go	0.00%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #886      +/-   ##
==========================================
+ Coverage   61.03%   62.97%   +1.93%     
==========================================
  Files          56       57       +1     
  Lines        5903     6012     +109     
==========================================
+ Hits         3603     3786     +183     
+ Misses       2143     2064      -79     
- Partials      157      162       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bwplotka]

FYI: Check an amazing https://pkg.go.dev/golang.org/x/perf/cmd/benchstat tool for easier comparision of those results 🤗

Also you can use the following command for (1) saving to file (2) do enough repetition for benchstat probabilistic significance check to work (reliability), (3) to customize duration (1s is by default) , (4) to customize how many CPUs you want to use:

export bench=bench1 && go test \
    -run '^$' -bench '^BenchmarkIdentifier' \
    -benchtime 10s -count 6 -cpu 2 -benchmem \
    -memprofile=${bench}.mem.pprof -cpuprofile=${bench}.cpu.pprof \
  | tee ${bench}.txt

[bwplotka]

I don't have context on other parts of Otel collector code, but storing those in maps is generally not ideal for efficiency. Especially as those attributes are short enough to be iterated to find items you are looking for etc

This is why Prometheus primary labels code started as array and now it's tightly optimized interned and encoded string 🙈 https://github.com/prometheus/prometheus/blob/main/model/labels/labels_dedupelabels.go#L27

Just ideas for future if even applicable.

[bwplotka]

This is a solid start 👍🏽 Good work!

[dashpole]

added delimiters to make sure we don't get collisions between the different maps