Hacker News Reader

Comment by ecopoesis | Hacker News Reader

A million requests per second in Python, if you write all the code that does work in C.

You use it in Python. OP writes the C code for you.

As soon as you do anything else than return 'hello, world' in Python, your performance will drop dramatically. Just parsing a tiny JSON string a million times in python takes 7s (i5-2467M).

  $ time python -c "import json;
  for i in range(0, 1000000):
      json.loads('{\"data\": \"hello, world\"}')"

  real	0m7.055s
  user	0m6.951s
  sys	0m0.063s

It is no use optimising one part of a system if other parts take 99% of the time.

As it stands, the title of the post would be more accurately stated as 'million requests per second with C'.

FWIW, you can use ujson, which is much faster than the builtin stdlib json parser as well. I've seen some larger webapps in python do just this.

json:

    $ time python -c "import json;
    for i in range(0, 1000000):
          json.loads('{\"data\": \"hello, world\"}')"

    real	0m3.142s
    user	0m3.098s
    sys	0m0.034s

ujson

    $ time python -c "import ujson as json;
    for i in range(0, 1000000):
          json.loads('{\"data\": \"hello, world\"}')"

    real	0m0.444s
    user	0m0.399s
    sys	0m0.028s

Like with many things, it's a question of being smart about your choice of libraries.

Also, just because I was curious:

    $ time pypy -c "import json;
    for i in range(0, 1000000):
          json.loads('{\"data\": \"hello, world\"}')"

    real	0m0.422s
    user	0m0.386s
    sys	0m0.032s

If you use a C library and have hardly any Python code running it obviously goes faster. That misses the point I was making. At some point you'll have to write some Python code, and that will lower the number of requests dramatically. If you're talking millions of requests per second, you will have maybe 10s of thousands of CPU cycles per request. That doesn't buy you much in Python. Case in point: parsing a trivial json string in Python eats up your budget.

Of course, most websites aren't dealing with anywhere near that volume of requests, and then Python is fine.

Pypy is interesting, but last time I looked it didn't support C extensions and hence can't be used with the poster's library.

time python -c "import ujson; for i in xrange(0, 1000000): json.loads('{\"data\": \"hello, world\"}')"

0.291s

Well, that was an easy way to reduce 7 seconds to 0.3.

You might want to use the timeit module instead:

  $ python -m timeit --setup 'import json' "json.loads('{\"data\": \"hello, world\"}')"
  100000 loops, best of 3: 9.34 usec per loop

And then you get all the fun of debugging clang internals when things go wrong.

And as little actual python code in the routes as possible...

Funny that people bother to do this when there's plenty of other languages that will handle the load without issues.

Java Golang Rust Scala Maybe C# C

The semi famous techempower benchmarks shows several popular frameworks that can handle over 1mill/sec

pretty sure Phoenix with Elixir can handle that load too.

Elixir/Erlang is really slow. The strengths of Erlang/Elixir is easy multithreading, easy distributed computing (just as easy as local multithreading), easy fault tolerance and easy low latency. If you have to do something single threaded and performance critical then you do the same thing in Erlang/Elixir as you do in Python: FFI to native code.

I wasn't implying Elixir/Erlang was fast.

My statement was that it could handle the load. Taken into account the fact that the language is built around concurrency, it would handle the load within multiple threads without a hitch.

By using C for the hot paths (e.g. HTTP parsing) and python for business logic you can get code that is both faster and easier to maintain than, say, Java.

HTTP is not the bottleneck of anyone's application. Any interesting business logic will almost certainly cause Python to perform worse than Java, Go, etc. I'm also skeptical of the maintainability claim, because I write Python professionally.

Writing web code in c isn't my idea of easy to maintain

If you used this framework you wouldn't be maintaining web code in C.

That's why you write in Python and use this library

The Python part would slow the whole thing down again.

Anything not written in assembly would slow things down. But very few developers write web apps in assembly these days.

Reply on news.ycombinator.com