bigpages and SGA size

[skt_skt]

I have a RHEL AS 2.1 with the below kernel

title Red Hat Linux (2.4.9-e.71smp)
root (hd0,0)
kernel /vmlinuz-2.4.9-e.71smp ro root=/dev/sda3 bigpages=1075MB usb-handoff
initrd /initrd-2.4.9-e.71smp.img

Also a oracle DB with the almost 1.5GB shared memory usage

------ Shared Memory Segments --------
key shmid owner perms bytes nattch status
0x17dc0028 327687 oracle 640 1598029824 1165

Not much memory free on the system. Half is dedicted to SGA.

total used free shared buffers cached
Mem: 3256 3251 5 882 142 346
-/+ buffers/cache: 2761 494
Swap: 4095 948 3147
Total: 7352 4199 3152


we obsere that the system gets hung while the oracle is getting patched and nothing works further even from the console.

"For whatever reason it looks when Oracle PID 26047 was running in that "prealloc" loop in shm_enable_bigpages(), it failed to get all of what it requested, so it started giving back what bigpages it had pre-allocated -- but in doing so it found an allocated bigpage page address that was illegal/impossible for a bigpage start."

[ sysctl -a|grep kernel.shm-use-bigpages
kernel.shm-use-bigpages = 2

I am trying to understand the relation between the big pages and SGA size if there is any..

[Steven E. Protter]

Shalom,

I know of no relationship between big pages, performance and SGA size. If the SGA size is too much for your system get more RAM or have the DBA reduce the SGA.

SEP

[skt_skt]

what would be the "HIGH MEM" available for a 3GB memory system? How is the calculation.