I'd also add that it's debatable whether using gunicorn's 'jsonify' is any better than json.dumps(). They had to add a shim in to make jsonify work. jsonify adds the correct header for json, whereas json.dumps doesn't. I'd expect the database performance to take a hit in this setup because of their shim, but not by much.
On Friday, April 5, 2013 12:41:47 PM UTC-7, Derek wrote: > > I've said in the past that they were unfair to Ruby because they were > using their ruby server in dev mode which recompiles code on every access. > > This time, I looked more closely at their 'Flask' setup. They are running > Flask via gunicorn, so you'd think all is good. However, they aren't > setting the 'worker type'. That means it's the default worker type, which > is 'sync'. In other words, it's the same as vanilla python. They should > have set it up with 'gevent' or 'eventlet'. And what do you know, they are > running it in development mode as well. Flag 'preload' needs to be True in > order for it to not recompile on every access. > > So, their setup as it is now, can handle exactly 8 requests at once, and > it recompiles code at every run. No wonder it's slow. Nobody in their right > mind would deploy their app that way. > With my changed setup, they could handle theoretically 8000 requests at > once, and it does not recompile at every run. > > I've submitted a 'pull request'. > > > On Friday, April 5, 2013 11:43:42 AM UTC-7, David Marko wrote: >> >> http://www.techempower.com/blog/2013/04/05/frameworks-round-2/ >> > -- --- You received this message because you are subscribed to the Google Groups "web2py-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to web2py+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
