zimbra busy with futex timeouts

[giray]

Hi All,
RHEL5.2 with Zimbra NE 5.0.11. The server is relatively busy with close to a load of 0.75 - 1 with 20 users seems a bit high. We also seem to need to restart zimbra every week.

When I strace the java process I see half of its time its in futex and
half of those it returns an error (ETIMEDOUT (Connection timed out))

% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
49.01 380.403215 11396 33380 14815 futex
45.40 352.331694 59126 5959 22 poll
4.13 32.056128 572431 56 accept

The box has 2 Gigs or RAM, swap is almost never used,
runs on VMware ESX with 2659Mhz CPU allocated.

any ideas, pointers and help welcome