Concurrency models (was: Timer)

[Ian Bicking]

On Monday, October 27, 2003, at 01:53 PM, Alan Kennedy wrote:
> I think threads are really a useful abstraction for programmers who
> are learning to deal with developing servers for the first time. And
> threads fulfill their function very well, to a point.
>
> But they have fundamental scalability problems: The C10K problem
> illustrates this well: Surely it should be possible for a modern
> computer, with giga-everything (!), to serve 10,000 clients
> simultaneously?

Fundamental scalability problems hit very few people.  And in fact 
there's nothing fundamental about them -- the particular of the 
interpreter and your concurrency model are only a very small part of 
the performance of a complex application.  Using threads or async and 
heavy caching, maybe you could outperform by two a multi-process model, 
making non-SMP scalability unimportant.  Or using multiple processes, 
you might be able to more simply migrate applications to other 
computers.  Real programs have a wide variety of constraints.

I'm unimpressed with most theoretical performance measurements.  O(n) 
analysis is, for most problems, meaningless, since it ignores 
non-scaling costs that can be very important.  SMP scalability is 
another concern I'm unimpressed with -- maybe you can get four times 
the performance with an SMP system, but so what?  It only counts if 
it's processor-bound, which isn't a lot of applications, and four times 
the performance isn't scalability, it just means you can wait a little 
longer before you partition or parallelize the task for separate 
servers.

--
Ian Bicking | ianb at colorstudy.com | http://blog.ianbicking.org