Il giorno 13/dic/2010, alle ore 18.46, [email protected] ha scritto:

> Hello Roberto,
> 
> 1. if I will create 1000 elements it will pre allocate 1000 * 65K?

No it will pre-allocate 4096*1000 as only the space for keys is malloced and 
zeroed
(but i have already find a way to avoid this reducing the pre-allocation to 
2*1000).


The space for values is mmaped but untouched until there is a real value on it.

> 2. how good idea to use it for few millions of items? or it's better
> to go with memcached then?

if the optimization algorithm does it work as i hope, the O(n) impact should be 
very low and so
you can put a lot of values in it. As the api is independent of the 
implementation there is a lot of space for
every kind of optimizations.


> 3. get/set/del operations are asyn or they are atomic?

not atomic for sure.

if you are in async mode the locking for write/del is a noop, otherwise it came 
in


So i prefer to define it as opportunistic-asynced and opportunistic-thread-safe 
:)

--
Roberto De Ioris
http://unbit.it

_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi

Reply via email to