my onstat -u, on IDS 7.3 at Solaris 7

Userthreads
address flags sessid user tty wait tout locks nreads nwrites
149953a8 S--PR-- 246 smp WSERVER1 15174938 0 1 29 32
14995894 S--PR-- 248 smp WSERVER1 15174938 0 1 20 32
14995d80 Y--P--- 485 smp WSERVER4 1526ebf0 0 1 0 0
1499626c S--PR-- 609 smp WSERVER4 15174938 0 1 11 0
14996758 S--PR-- 281 smp WSERVER1 15174938 0 1 18 0
14996c44 S--PR-- 614 smp WSERVER4 15174938 0 1 13 0
14997130 Y--P--- 254 smp WSERVER1 14b4be28 0 1 10 0
1499761c S--PR-- 615 smp WSERVER4 15174938 0 1 5 0

what actually wrong with this machine?
the performance is really bad.

Hi,

You can extend the information sending but data like memory, amount of
processors, onconfig, disk layout if these using raw-devices or not.

Gustavo.

You haven't included enough information. Yes, you're waiting on latches (Mutexes in Unix) but before you look at Informix look at the operating system.

Check release documentation and find out what the recommend kernel values are for your release. Check these against actual values - and remember, the release documents assume this is a dedicated Informix server. If something else runs on this server you need to factor.

Is this a new problem or ongoing? Fess up - what changed?

Check onstat -s and see who's hold in the latches. If no one is and everyone's still waiting, you're short at the operating system level.

How many users, how many CPU's (defined vs. actual) - and once again, are you the only player on this box?