I don't think that the MRG low latency will every be moved into
HTCondor core. But there are some things that can be done to reduce the latency of HTCondor running jobs now. Can you be more specific as to your needs ? are you looking for latencies on the order of a few seconds? or does it need to be more like a few milliseconds? Do you still want (or need) to have a the SCHEDD hold and manage the job queue? The current HTCondor has two way for STARTDs to be handed jobs directly without going through the SCHEDD, and thus not dependent on the negotiation cycle. One way is to use work fetch hooks on the STARTD. http://research.cs.wisc.edu/htcondor/manual/v8.3/4_4Hooks.html#49806 tj On 8/26/2015 8:33 PM, Rita wrote:
|