We are running tests on a large database that has been running very slowly. We are trying to speed it up. AIX is 4.3.3 and it is an Oracle database 8.1.7. We've got lots of memory (8 GB) and 6 CPUS @ 500 MHz. Paging is about 5%.
When the DBA started running his tests, I started topas to watch disk activity in particular. Things were going along normally. Then, all of a sudden, all of the disks (I had 12 of them displayed) went to 0% busy, but topas also said they were writing thousands of KBPS. This went on for about 5 minutes and even continued after the DBA stopped his test.
I suspect that the syncd may have kicked in, but I did not see that process in the process list in topas. I changed the vmtune maxrandwrt parameter from the default of 0 to 128 and we got better performance. What would a "typical" value be for this parameter? And has anyone ever seen disks being 0% busy but writing like crazy for about 3 to 5 minutes?
When the DBA started running his tests, I started topas to watch disk activity in particular. Things were going along normally. Then, all of a sudden, all of the disks (I had 12 of them displayed) went to 0% busy, but topas also said they were writing thousands of KBPS. This went on for about 5 minutes and even continued after the DBA stopped his test.
I suspect that the syncd may have kicked in, but I did not see that process in the process list in topas. I changed the vmtune maxrandwrt parameter from the default of 0 to 128 and we got better performance. What would a "typical" value be for this parameter? And has anyone ever seen disks being 0% busy but writing like crazy for about 3 to 5 minutes?