[zeromq-dev] Thinking out loud ...

Andrew Hume andrew at research.att.com
Fri Jun 24 21:54:01 CEST 2011


because the desired granularity (single job) is so small
compared to all teh buffering involved, this kind of thing is, i think,
best handled by REQ/REP initiated from the worker.

i have handled similiar things before in this way, and even had
the central job dispenser solve for optimal job distitubtion
in between job requests.

	andrew

On Jun 24, 2011, at 12:45 PM, Martin Sustrik wrote:

> On 06/24/2011 09:20 PM, Henry Baragar wrote:
> 
>> Yes, I have a couple of workload distibution applications with highly
>> asymmertric workloads (see my message from from 2011-04-04 "Re: LRU
>> broker queue in intuitive way on 3.0
>> <http://lists.zeromq.org/pipermail/zeromq-dev/2011-April/010352.html>").
>> I can't figure out how to configure 0MQ in such a way that I can add
>> additional workers after the application has started, because all the
>> messages already have been sent (and queued) to other workers.
> 
> OK. AFAIU what would solve your problem is to strictly limit the number 
> of messages in flight to a particular peer, ie. honour HWM exactly 
> without having excess messages in TCP buffers, right?
> 
> Martin
> 
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev


------------------
Andrew Hume  (best -> Telework) +1 623-551-2845
andrew at research.att.com  (Work) +1 973-236-2014
AT&T Labs - Research; member of USENIX and LOPSA




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110624/7c031509/attachment.htm>


More information about the zeromq-dev mailing list