[vworld-tech] Python for the backend

[ceo]

Alex Chacha wrote:
 > and then I did a small 10 client load test and saw that it just did not
 > scale well at all.  Which was disappointing.  I then tried to write a

When I saw that, I was disappointed - didn't think it was that bad, but 
I'd heard it was bad...

 > simple java based implementation and that was a bit better but delays
 > were huge when more than 15 clients were connected at the same time.

...but now I *know* you're doing something very, very wrong, or you're 
running something incredible (like a Euler-esque physics engine, or a 
distributed raytracer), because java scales fairly effortlessly to 500 
clients these days (i.e. use the sun tutorials aimed at newcomers to the 
language, and you should see 500 clients straight away).

So ... if you'd like some advice, post what you're doing (in english, 
please - if you post 300 lines of code I'm not going to read it, 
although code snippets are great at 10 or so lines) and I'll see if I 
can spot any obvious improvements.

 > you some ideas about what to expect in the long run.  If you plan on
 > having maximum of 20 people simultaneous you can use python or java and
 > try not to have 1 thread per socket or you will hit the dreaded
 > "more-time-spent-on-context-switching-than-execution" problem.  If you

You start to notice that at 75-150 or so, usually. Hitting that any 
earlier is ... unusual.

 > bet.  With java and python you can always use more machines and put a

Entirely unrelated, but ... I'm about to use java to front-end a system 
that's got to run 50 million hits per hour, and I have no concerns about 
it managing that (based on experience, not hype). I'll talk more on that 
in another post, in case anyone's interested. But the long and the short 
of it is that with java you should be hitting exactly the same 
performance for a server as you get with C++ (though I'm not qualified 
to make such statements w.r.t. Python).

Adam