Core to workload functionality
Hi,
Can someone please help me understand what the "core to workload functionality" mentioned in the process analysis section, really means..
++++++++++++
The process analysis identifies processes/LWPs that might cause a singlethreaded bottleneck. The relationship of any process/LWP consuming more than 20% of a core to workload functionality must be understood as it might become a bottleneck on Sun Fire T2000/T1000 servers. Consuming CPU resources is a good characteristic when doing throughput computing. When processes or threads have dependencies on a single process or thread that consumes execessively large quantities of compute resources, analysis of this potential bottleneck is required before migration.
+++++++++++++++++++++
Thanks in advance..
DG
[805 byte] By [
danugg] at [2007-11-26 11:25:08]
# 1
The basic advantage of T1000/T2000 processors over standard ones is
based on the possibility to eliminate excessive idle time of processes/LWP
by the possibility to switch to another HW thread of the core when the
previous is idle.
If the mentioned above "another HW thread" is dependent on the previous
one, then its execution cannot be continued and the initial (idling)
process/LWP becomes effective cause of the observed bottleneck.
In a more favorable scenario, where the first idling thread is not impacting
the next thread(s), the processor is being used more efficiently by spending
its activity on threads which can progress without waiting.
This is the reason why it is important to "understand" the relationship
of the process/LWP consuming a substantial executing time to the
other components of the workload.
zenon, PAE
zenon at 2007-7-7 3:40:35 >
