Revised Round 5 of TechEmpower benchmarks: Revel vs plain Go

[Brian Hauer]

Hi all,

We've posted a minor revision of Round 5 of our framework benchmarks.  Notably, this includes a test of Revel on i7, which had been unintentionally omitted in the original Round 5 post.  The plain Go tests were not re-run for this minor update.  They will be re-run again for Round 6.  Also, we didn't re-run any Windows tests, so the Go + Windows combination that didn't successfully run its database tests in Round 5 remains unchanged.

Basically, we just wanted to correct a couple easy-to-fix omissions in Round 5.

The full results:

http://www.techempower.com/benchmarks/#section=data-r5

Just Go:

http://www.techempower.com/benchmarks/#section=data-r5&p=3k

The most interesting things to my eye are:

1. Revel has a low "framework penalty" as compared to other frameworks versus their platforms. 
2. Revel shows a higher performance on the 20 query test on i7 than plain Go.  This could be related to the manner by which the 20 queries are delivered to the MySQL driver.  I'd be interested in hearing any Go experts' thoughts on that and would be happy to revisit the plain-Go test's implementation if there is room for improvement.

Revel test code: https://github.com/TechEmpower/FrameworkBenchmarks/blob/master/revel/src/benchmark/app/controllers/app.go

Plain Go test code: https://github.com/TechEmpower/FrameworkBenchmarks/blob/master/go/src/hello/hello.go

Thanks again to everyone who has helped with this project.  The Round 6 ETA is June 20.

[robfig]

Boy, that's embarrassing.  It's not picking a new random row for each query, so it repeatedly loads the same row.  Preparing a pull request... 

[Brian Hauer]

Oh no!  Well, we'll get that re-run when you get the pull request submitted.  Thanks for looking into it for us!

[robfig]

https://github.com/TechEmpower/FrameworkBenchmarks/pull/313

The implementation is exactly the same, so it is certainly a red flag that it came out ahead :p.  I would not expect it to, although maybe I will put a bit more effort into tuning ahead of Round 6.  (Low hanging fruit exists, like right now it sets and parses 3 different cookies for every request). 

Sorry again for the incorrect results.