RE: AIX Concurrent I/O for Oracle archive logs
- From: "Allen, Brandon" <Brandon.Allen@xxxxxxxxxxx>
- To: "dbvision@xxxxxxxxxxxx" <dbvision@xxxxxxxxxxxx>, "oracle-l@xxxxxxxxxxxxx" <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 27 May 2010 10:14:13 -0700
Yes, I wouldn't expect much of a performance boost from using CIO on archived
logs, but that wasn't my point - I'm only considering putting them on CIO in
order to prevent them from flooding my AIX buffer cache when a log backup runs
- below is a real life example of what happens on our system during an
RMAN/Netbackup archive log backup - notice free memory drops from 200513 pages
to almost 0 and paging kicks into high gear. Once that happens, the system gets
almost completely frozen for about 5 minutes before it clears itself up again.
kthr memory page faults
cpu time
------- --------------------- ------------------------------------
------------------ ----------- --------
r b avm fre re pi po fr sr cy in sy
cs us sy id wa hr mi se
2 0 5788490 200513 0 7 0 0 0 0 2368 12733
5469 35 10 54 1 10:16:55
1 1 5782725 188188 0 34 0 0 0 0 3782 27734
8698 38 7 52 3 10:17:05
1 0 5781430 168965 0 4 0 0 0 0 3199 24317
6279 34 7 57 2 10:17:15
2 0 5781582 148703 0 12 0 0 0 0 2734 21061
6227 40 7 51 2 10:17:25
1 1 5777700 131101 0 29 0 0 0 0 2372 11121
4857 30 8 59 3 10:17:35
1 1 5777675 101893 0 2 0 0 0 0 3451 11643
6567 30 7 56 6 10:17:45
2 0 5780391 76652 0 0 0 0 0 0 2692 18970
6273 33 7 58 2 10:17:55
3 1 5793782 39854 0 21 0 0 0 0 3374 34886
9554 52 12 33 3 10:18:05
8 1 5798319 19155 0 9 0 0 0 0 3948 103127
29678 72 19 7 2 10:18:15
4 4 5803921 23 0 37 211 567 1375 0 3427 70698
26765 65 10 11 14 10:18:25
1 8 5809971 19 0 9 1707 1996 4756 0 2411 61226
29619 38 12 18 32 10:18:35
2 16 5813055 5 0 16 1647 2212 17302 0 2828 55202
27428 37 12 15 36 10:18:45
I agree - the 64K pages may be a little too close to the bleeding edge. I've
also decided to hold off on implementing them for now - in fact we just had the
discussion here last week for another environment where we're currently
migrating a 10.2.0.4 database from AIX 5.3 to a new AIX 6.1 server and we
considered switching from 16M pages to 64K at the same time, but decided to
just stick with 16M for now since they're already configured and have been
working well for years.
Regards,
Brandon
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On
Behalf Of Nuno Souto
But let's consider the IO that archived redo logs have:
almost always sequential
. . .
I wouldn't worry about the type of IO.
Yes, I've seen the 64K notes.
. . .
I'll let someone
else iron out the bugs...
Privileged/Confidential Information may be contained in this message or
attachments hereto. Please advise immediately if you or your employer do not
consent to Internet email for messages of this kind. Opinions, conclusions and
other information in this message that do not relate to the official business
of this company shall be understood as neither given nor endorsed by it.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
