Benchmark about bjoern

[jonashaag]

bjoern is a server, not a framework.

[vmandke]

Is there something wrong with the way I have run Bjoern ? Benchmarking Gunicorn vs Bjoern with hello-world POST request.

http://stackoverflow.com/questions/41531707/bjoern-v-s-gunicorn-post-requests?noredirect=1#comment70269241_41531707

[jonashaag]

What tool are you using for the benchmarks? Are you running them from the same host that the servers run on? Why are you using multiple workers for bjoern?

Don't expect huge differences in speed. Time spent in the web server will be roughly 5%, the other 95% is spent in your application code. For example if you make a single database query in your application code then the choice of web server simply doesn't matter anymore for performance.

[vmandke]

@jonashaag I am using Tsung for benchmarking
No, the tsung node and server nodes are different, however both have same configuration. Ubuntu 12.04.5 LTS (GNU/Linux 3.2.0-115-virtual x86_64)

Using two workers as there are 2 cores on the node.
The flask app simply returns the JSON it got from the POST request.

[jonashaag]

I guess you're simply measuring network performance then, however I don't know what's wrong with your setup that you get such different request rates. I suggest you start your benchmarking with a simple hello world app.

[vmandke]

@jonashaag

the flask app I am using just does this:

from flask import Flask, request, jsonify

app = Flask(__name__)


@app.route('/suggest/', methods=['POST'])
def hello():
    content = request.get_json()
    return jsonify(**content), 200

[jonashaag]

That's not hello world and you're not running the tests from the same node. Start with the simplest possible setup.