Tornado benchmark

[Dith3r]

1. Usage of tornado.web.asynchronous instead of gen.coroutine
2. You don't setup backlog for tornado
3. You can pass data to be written to finish instead of write + finish.

[Dith3r]

4. Tornado is generating ETag.

[Aeron]

Also, Tornado can use AsyncIO as well, moreover UVLoop can be used as a replacement for standard AsyncIO’s loop. This will demonstrate very different results.

I think it’s only fair to add such cases in the comparison. Pure AsyncIO at least. Especially, since these options are far from some exotic single-purpose combinations, they’re very common. Maybe even more popular these days than pre-AsyncIO implementation, which you’re comparing with.

[frnkvieira]

Do you mind doing a pull request modifying the files at (https://github.com/vibora-io/benchmarks/tree/master/frameworks/tornado_)? It should be quite straightforward and as tornado users (which I'm not), you guys are much better than me on this task. Anyway, I appreciate your comments, thanks!