vBulletin Search Engine Optimization
| |||||||
| Register | FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| google account wrote: > > Hello... > > My OpenServer 5.0.6 system seems to be suferring random go slows, and > I am getting data out of sar suggests strongly that it is a disk > system bottleneck and I should increase my NBUF to cope better with > it... My big issue is that I really have absolutley no idea how to > increase the NBUF figure, nor do I know how much to increase it by so > as not to run myself low of memory in the system > > Can someone let me know...? Or point me to instructions? > > Just incase it make a difference, the system itself is an Intel Mobo > with dual P3 Xeon processors and 1Gig of physical memory. I beleive > it has been configured to have 3gig swap. The disk system is a SCSI > RAID5 + hotspare thing, with 4 10000rpm 36g scsi drives... > > Thanks > > googleboy. On a different angle, if your raid controller support RAID 1+0, then reconfigure your hard drives. You will have to unload/reload with a supertar. I have seen RAID 5 causing a bottleneck, normally with a database and numerous small writes. The RAID 1+0 has better performance in the same test. With a three drive RAID 5 and one spare you can sustain a single hard drive failure without loss of data ( performance will be hurt even more during the rebuild ) but a second failure will cause a total volume loss. With RAID 1+0 you can have a single drive failure with no loss of data ( performance degredation will occur when the failed drive is replaced, but its less noticable then with RAID 5 ), and it is possible that a second hard drive can fail without the volume going down. A third failure before any drive replacement will always kill it. So RAID 1+0 is no less secure than RAID 5, has better performance and you already own the four drives. Mike -- Michael Brown The Kingsway Group |
| Thread Tools | |
| Display Modes | |
|
|