Seems like there was a misunderstanding here.

Note that there are different iterations of the tests, and we are talking about 
the ones published as " **round 17** ".

Also note that there are 12 result tab combinations: 6 tests times 2 test 
environments. **Nim participated in four of the twelve results** (with its 
fastest framework coming in #1, #20, #10, and #14), and **Nim didn 't 
participate in eight of the results**.

Nim hasn't yet submitted code for most of the tests on TechEmpower. I keep this 
thread active to encourage more qualified people than myself to look into this.

[My post](https://forum.nim-lang.org/t/1152#27100) had emphasized the good 
news, because it was indeed a big achievement. I'm sorry if that was the cause 
of the confusion.

I think good Web benchmarks results are very important for Nim's popularity 
advancement. Every major business has a Web-site, and internal apps that are 
increasingly Web-based. Many users of scripting languages start considering a 
statically-typed compiled language when they realize their current code doesn't 
scale, and Nim should be an especially good candidate for people coming from 
Python.

> the memory is the same for all

The **total** memory is the same for all frameworks: 32g in the "physical" test 
environment (Dell R440) and 14g in "cloud" test environment (Azure D3v2).

But different tests **used** a different amount of memory. Which IMHO an 
important thing to measure. That's why I

Let's say I have a Web service that needs fairly little storage, bandwidth, and 
CPU; and it has the same memory requirements for decent performance as the 
memory usage in TechEmpower's "Plain Text" test on Azure. If I use a memory 
framework that used <2GB RAM (ex. go fasthttp), I only need $7/month for [VPS 
hosting](http://ramnode.com/vps.php). But if I use a memory-hungry framework 
(ex. java comsat-servlet-undertow), I would pay $28/month for the extra RAM. 
Very big difference!

> lots of mem can only somewhat minimize java's disadvantage.

That's exactly my point. Lots of RAM hides the fact that Java would be slower 
on a cheaper server / cloud instance.

> I love Nim and consider it to be the (almost) perfect language (that just 
> needs a bit more growing up and mature).

Amen.

* * *

In other news, a less popular anecdote [I mentioned 
here](https://forum.nim-lang.org/t/1152#22964) 
([tbrand/which_is_the_fastest](https://github.com/the-benchmarker/web-frameworks)
 now redirects to 
[the-benchmarker/web-frameworks](https://github.com/the-benchmarker/web-frameworks))
 has been updated, and it seems Nim there is out of tune. It has two results:

  * [Latency](https://github.com/the-benchmarker/web-frameworks#latency) \- 
Jester's 6.82ms average isn't bad, but the table being sorted by the 50th 
percentile makes it look worse.
  * [Requests Per 
Second](https://github.com/the-benchmarker/web-frameworks#requests-per-seconds) 
\- Jester has the highest throughput (315.19 MB), but comes 7th in RPS.


Reply via email to