Performance of Flask, Tornado, GEvent, and their combinations.md

Performance of Flask, Tornado, GEvent, and their combinations

Wensheng Wang, 10/1/11

Source: http://blog.wensheng.org/2011/10/performance-of-flask-tornado-gevent-and.html

When choosing a web framework, I pretty much have eyes set on Tornado. But I heard good things about Flask and Gevent. So I tested the performance of each and combinations of the three. I chose something just a little more advanced than a "Hello World" program to write - one that use templates. Here are the codes:

1, Pure Flask (pure_flask.py)

from flask import Flask, render_template
app = Flask(__name__)

@app.route('/')
def main_handler():
   return render_template('main_j2.html', messages="whatever",title="home")

if __name__ == '__main__':
   app.run(port=8888, debug=False)

2, Pure Tornado (pure_tornado.py)

import os.path
import tornado.httpserver
import tornado.ioloop
import tornado.web

class MainHandler(tornado.web.RequestHandler):
   def get(self):
      self.render('main.html', page_title="", body_id="", messages="whatever",title="home")

settings = {
   "static_path":os.path.join(os.path.dirname(__file__),'static'),
   "template_path":"templates",
}
application = tornado.web.Application([
   (r"/", MainHandler),
], **settings)

if __name__ == "__main__":
   application.listen(8888)
   tornado.ioloop.IOLoop.instance().start()

3, Flask with Gevent (gevent_flask.py)

from gevent.wsgi import WSGIServer
from pure_flask import app

http_server = WSGIServer(('', 8888), app)
http_server.serve_forever()

4, Flask with Tornado (tornado_flask.py)

from tornado.wsgi import WSGIContainer
from tornado.httpserver import HTTPServer
from tornado.ioloop import IOLoop
from pure_flask import app

http_server = HTTPServer(WSGIContainer(app))
http_server.listen(8888)
IOLoop.instance().start()

5, Tornado with Gevent (gevent_tornado.py)

import tornado.wsgi
import gevent.wsgi
import pure_tornado

application = tornado.wsgi.WSGIApplication([
   (r"/", pure_tornado.MainHandler),
],**pure_tornado.settings)

if __name__ == "__main__":
   server = gevent.wsgi.WSGIServer(('', 8888), application)
   server.serve_forever()

I have 3 template files: main.html, layout.html, and form.html. main.html "extends" layout.html and "includes" form.html. The total size of templates is about 30kB.

The reason Flask and Tornado use different templates (main.html and main_j2.html) is that their template syntax is slightly different. Flask (jinja2) template use "{% endblock %}", "{% endfor %}", etc. while Tornado just use "{% end %}".

I tested performance (requests per second) using ApacheBench:

ab -n 1000 -c 4 http://localhost:8888/

and run it 5 times. The testing is done on a 6-year old dual-Opteron 254 server.

Here are the results:

pure_flask:      82  88 107 102  71
pure_tornado:   144 244 241 294 290
gevent_flask:   127 139 145 152 110
tornado_flask:  110  88  74  92 101
gevent_tornado: 328 555 177 273 153

Here are the averages:

pure_flask:      90
pure_tornado:   242
gevent_flask:   135
tornado_flask:   93
gevent_tornado: 297

As you can see, the Tornado implementation is significantly faster than Flask. Gevent makes Tornado faster, but not by a lot.

In the end, I like the straightforward style of Tornado and not the Flask way to write large project (using blueprints), So I sticks with Tornado.

gunicorn +1

uvicorn

FastAPI vs Tornado