Re: [HTCondor-users] Issue with 12 - 8 core machines bottlenecking

[Rich Pieri <ratinox@xxxxxxx>]

Stuart Wilson wrote:
> job on each HDD.  Is there a way to mimic this using condor?  Were
> afraid when we have condor run a job on those 8 and 12 core machines
> with one HDD will see a delay in performance.

Since you're running with single disks (no redundancy) I suggest
reconfiguring the drives as RAID 0 (stripe) sets. This will improve read
and write performance for all jobs regardless of the job submission system.

I also suggest adding more disks to each chassis. Two disks in a RAID 0
set have roughly double the I/O performance of a single disk. Four disks
in a RAID 0 set have roughly double the I/O performance of the 2-disk set.

Given an unlimited budget? SSDs are the way to go. They are expensive.
They are short-lived. They are very, very fast.

-- 
Rich Pieri <ratinox@xxxxxxx>
MIT Laboratory for Nuclear Science