The project

I recently started working on a cryptocurrency exchange aggregator. Basically I send out requests to a bunch of different exchanges and compare rates. This has to be made as fast as possible. In this post, I’ll some show some tweaks that I made in order to boost my performance significantly.

However keep in mind that I am not an expert (especially in go) and I am just sharing my findings from my own personal project.

Improvements

These improvements come in order of biggest improvement of runtime.

1. Using goroutines

In any Go program, goroutines are essential for speed. The biggest boost I made was by sending requests concurrently. Since I need to hit up 12 different exchanges, sending these requests at the same time dropped my runtime from around 24 seconds to just ~3.

Goroutines are amazing and extremely easy to use. You should include them wherever possible. But always be careful of Data Races

2. Upgrading the JSON Library

I swapped out encoding/json for github.com/json-iterator/go. jsoniter is a fast JSON processing library that works as a drop-in replacement for the standard library, so I didn’t have to change any code, just a library switch.

Benchmark Results

To measure the performance improvements, I ran some benchmarks comparing encoding/json and jsoniter. Here’s a summary of the results:

goos: linux
goarch: amd64
pkg: apiSpeedImprove
cpu: AMD Ryzen 5 7640U w/ Radeon 760M Graphics      
BenchmarkEncodingJSON-12          140383              7381 ns/op
BenchmarkJSONIter-12              974605              1217 ns/op
PASS
ok      apiSpeedImprove 3.216s

So, jsoniter is about 6 times faster than the standard library.

3. Reusing HTTP Handlers

I started reusing HTTP handlers instead of creating new ones for each request. By setting up a handler once and reusing it, I cut down on the overhead of making new handlers for each request.

Benchmark Results

Here are the results of the benchmarks comparing reused handlers versus creating new handlers for each request:

goos: linux
goarch: amd64
pkg: apiSpeedImprove/httpReuse
cpu: AMD Ryzen 5 7640U w/ Radeon 760M Graphics      
BenchmarkNewHttpClientEachRequest-12                3360            300058 ns/op
BenchmarkReuseHttpClient-12                         6470            175472 ns/op
PASS
ok      apiSpeedImprove/httpReuse       4.010s

Reusing HTTP handlers gave a solid performance boost compared to making a new handler for each request.

Conclusion

With these tweaks I managed to cut the time it took to gather all the info from 24 seconds initially to about 2 seconds. Pretty solid improvement!

If you are interested the code for my benchmarks, it is available here

If you enjoyed this post and want to support my work, you can donate here.