first time user with out of memory question

[Paul T]

--- Ivan Krstic <krstic at fas.harvard.edu> wrote:

> Paul T wrote:
> >  You're right. What is implemented in memcached
> should
> > be called "Lazy Garbage Collection". 
> 
> No, it shouldn't. Memcached doesn't do GC. It does
> LRU ejection.

 Unfortunately, as per google, such a thing exists
only in your head :

http://www.google.com/search?hl=en&lr=&q=%22LRU+ejection%22&btnG=Search

> >  Either :
> >  "it is so simple - it can not have bugs" or 
> >  "I just fixed the last bug in that part of code" 
> >  - it can not be *both* at the same time, right? 
> 
> It turns out that in various usage scenarios, the
> way a slab allocator
> allocates memory is counterintuitive if you don't
> prime it, as Steve
> mentioned, by sticking in an object for each size
> class that you care
> about. So it's not that the allocator is buggy or
> complex; it's that
> it's sometimes too simple for its own good.

 If you say so.
 
> > Unfortunately, memcached is an asynchronous
> > application (based on libevent). 
> 
> How is this unfortunate?

 For example, because one can not apply automated
tools, such as valgrind, to debug such an application.

 libevent is cool. I have recently completed a server
based on libevent, serving 30K hps. 

> > For that kind of
> > applications (timing/pattern/load/OS sesnitive) --
> "it
> > runs in my environment" does not mean that it
> would
> > work for a different setup.
> 
> That's a complete non-sequitur.

 Try placing memcached under 30K hps load and you
might begin understanding what I'm talking about. 

 Or you can try googling - there are plenty of good
stuff out there. - some is with numbers. 

 A nice entry point is :
http://www.kegel.com/c10k.html

Rgds.Paul.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com