On Dec 2, 2004, at 1:43 PM, Mark D. Rodriguez wrote:
Hi,
I have a Linux based ITSM server running RHAS 2.1 with ITSM 5.2.2 server code. The ITSM server is an IBM X345 w/ 2 proc., 3GB memory, disk storage is on FASTt. There are a mix of Windows, Novel, and Linux clients running ITSM 5.2.2. This is a new environment and we are still working out the kinks. The biggest issue is the Aggregate transfer rate on all of the machines is below 400KB/Sec. on a 100Mb ethernet connection. We have done FTP test that show us transfer rates more like what we would expect, i.e. approximately 10MB/sec.
I am wondering has anyone else seen this kind of performance problem? Were you able to solve it and if so how?
Well, you'll find many historic postings of such a question. The Aggregate data transfer rate is based upon session elapsed time, and so networking tends to be a relatively minor element in it. With mature file systems, the client internal processing time dominates (file system traversal seeking candidates being the major factor, as possibly impeded by virus detectors, etc.). Performing a Selective backup on a multi-GB file may be more illuminating, particularly in a mixed clients environment with a common TSM sever. Doing Query SEssion as normal backups run will give you a sense of delays, TSM accounting records provide numerics on wait times.
This boils down to classic systems analysis, where we have to inspect all the elements contributing to the whole to determine the time spent in each area. You can always fall back on TSM client tracing to enumerate the factors that it knows about. The TSM Performance Tuning Guide is helpful up to a point, but we have to have a sense of our system to know what tuning is appropriate.
Richard Sims
