[zeromq-dev] scalability of zeromq in terms fof queues and topics

Bhavin Turakhia bhavin.t at directi.com
Sat Apr 10 06:51:00 CEST 2010


Thanks for the links. We checked the code and I am kicking myself for not
realizing that a basic trie would be all that's needed :). We also noticed
the nice lil optimization to the trie data structure implementation to
reduce storage space requirements :)

- Bhavin

> -----Original Message-----
> From: Martin Sustrik [mailto:sustrik at 250bpm.com]
> Sent: Thursday, April 08, 2010 7:55 PM
> To: Bhavin Turakhia
> Cc: zeromq-dev at lists.zeromq.org
> Subject: Re: [zeromq-dev] scalability of zeromq in terms fof queues and
> topics
> 
> Martin Sustrik wrote:
> 
> > It's practical O(1), there's no hash table. It's basically an N-ary
> > search tree where first branching corresponds to first character of
> the
> > topic, second branchings correspond to second character etc. Thus you
> > never get hash conflicts.
> 
> Btw, here's the code. It's pretty simple:
> 
> http://github.com/sustrik/zeromq2/blob/master/src/prefix_tree.hpp
> http://github.com/sustrik/zeromq2/blob/master/src/prefix_tree.cpp
> 
> Have a look and give it a thought. If you can make it even faster,
> let's
> go for it!
> 
> Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 3074 bytes
Desc: not available
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20100410/adda49ec/attachment.bin>


More information about the zeromq-dev mailing list