OK, all the new ~poop~ is multi-threaded. Anyone know the scoop on LWP's (light weight processes), specifically under Solaris ?
We noticed a slow down in some environments upon upgrading to latest TSM 4.2.2.1 & tdp/r3 3.2.0.11 now that I'm bald from pulling out all my hair we notice that on a sun box with 32 GB memory & 32 processors we see a 27 session tdp/r3 backup action run with 40 LWP's and do OK BUT a sun box with 6 GB memory & 15 processors we see a 25 session tdp/r3 backup action run with just 7 LWP's and do less than expected... actually about the same throughput as we saw before if we just initiated 7 sessions (sigh) (find your "backint" process and use a "ps -lLf -p <backintpid>" to see the associated light weight processes) I have this into IBM also, to ask the developers if they do anything in the code to limit the number of LWP's but who knows how long that will take...??????? OH, so if you've upgraded lately to some of this multi-thread code and noticed a slow down (mainly if you use tsm client compression) it might be LWP related. Dwight E. Cook Software Application Engineer III Science Applications International Corporation 509 S. Boston Ave. Suite 220 Tulsa, Oklahoma 74103-4606 Office (918) 732-7109
