Is this an 11g or 10g PoC? I’m very interested!

bw.

I have not tested them myself. No, that URL wouldn’t hint an advantage for Oracle Database. On the other hand, look at page 16 of that report (SPEC int). Now that… is more in line with a Oracle. Remember, Oracle is a load and store, integer-rich workload.

I wish they would have done that testing with Linux though.

That single-instance PDF has apparently been pulled down from Oracle’s site… there seems to be an 11g version up now:

http://www.oracle.com/technology/products/database/clusterware/pdf/SI_DB_Failover_11g.pdf

Have you had an opportunity to compare Oracle performance of the AMD 23xx “Barcelona” and Intel E54xx “Harpertown” now that these CPUs are in shipping servers?

Is the throughput measurement shown here a meaningful indicator of an AMD advantage for some aspects of running 10gR2 or 11g? http://www.hardware.info/en-UK/articles/amdnZWppZGWa/New_quad_core_server_CPUs_AMD_Barcelona_vs_Intel_Harpertown/18

It turns out that mention of a particular setting in a Validated Configuration doesn’t imply a support requirement. That aside, I really can’t comment on anything NUMA related in RHEL3 since I abandoned that long before I even got that DL585…which I no longer have since it is in an HP lab and I’ve moved on to Oracle…

It turns out that mention of a particular setting in a Validated Configuration doesn’t imply a support requirement. That aside, I really can’t comment on anything NUMA related in RHEL3 since I abandoned that long before I even got that DL585…which I no longer have since it is in an HP lab and I’ve moved on to Oracle…

I’m curious, have you done any testing or have any insights into vm.numa_memory_allocator? There is little info available on this tunable that I’ve found so far, and before I start sifting through kernel source, I wanted to throw this out there. This was added to Update 7 of RHEL 3 (yes, I know, a bit dated at this point), but since Oracle recommends disabling NUMA on DL 585 gear, and I’m using that with RHEL 3 U8+, I’m curious about no NUMA vs. leaving it enabled in the BIOS and using this tunable. I’m wondering if it helps avoid the cyclops scenario…

Thanks,

-charles.

I just about censored out your comment, but then I got to thinking that you might actually believe NFS over GbE is limited to 33Mb/s. You couldn’t be further from the truth. How do you suppose NFS eats up 967Mb? I think you just mistook your nomenclature. Nonetheless, I have provided countless evidence that Oracle over NFS can easily drive throughput up to line speed (~114MB/s-118MB/s)

…there is nothing misleading about that.

Read:

http://h71028.www7.hp.com/enterprise/downloads/15650%20NAS%20Oracle%20WP%204A2.pdf

so 114Mb/s is a bit missleading

sqlplus -silent < /tmp/$$.1
a/a@${DBNAME}
exit;
EOF
egrep ‘ORA-121|ORA-01034′ /tmp/$$.1 > /dev/null
if [[ $? = 0 ]]
then
mailx -s ” ${DBNAME} is not accessible - db is down” ${MAILLIST} /dev/null
if [[ $? = 0 ]]
then
rm /tmp/$$.1
exit 16
else
mailx -s ” ${DBNAME} is not accessible - db is down” ${MAILLIST} < /tmp/$$.1
fi
fi

rm /tmp/$$.1

exit

Any ideas?
Eamon.

I have a clariion cx700. Each shelf will hold 15 drives. If I have 4 shelfs, 60 spindles, how do I lay them out for the best performance?

I think it is a no brainer to let the san do the mirroring, so I can present 30 small logical drives to ASM.

Or, I could mirror and strip two shelves presenting two huge drives to ASM.

Or, a little of both, I could mirror and strip six drives from two shelves at a time to present 10 medium drives to ASM.

In general, which do you recommend?

Thanks,
Reece

We do also see average Log File Sync waits that are much larger than Log File Parallel write when there are a large number of sessions commiting and CPU utilisation is 20% - 30%

It is possible for a session commiting to issue its commit a fraction after the current write in is progress, therefore presumably it will have to wait for the current write to complete and the next which contains is own commit marker. It is therefore not unreasable to see a Log File Sync wait which is almost double the Log File Parallel write time which will affect the average.

Thanks

Even so, sqlplus would receive an “enter” if the statement was followed by a blank line.
If your “alter session” would have been directly followed by the select statement, sqlplus would have given an error message:

sys@ORCL> host type c:\little_things.sql
alter session set nls_date_format=’DD/MM/YYYY HH24:MI:SS’

select sysdate from dual;

alter session set nls_date_format=’DD/MM/YYYY HH24:MI:SS’;

select sysdate from dual;

alter session set nls_date_format=’DD/MM/YYYY’
select sysdate from dual;

sys@ORCL> @c:\little_things.sql
sys@ORCL> alter session set nls_date_format=’DD/MM/YYYY HH24:MI:SS’
2
sys@ORCL> select sysdate from dual;

SYSDATE
———-
29/03/2008

sys@ORCL>
sys@ORCL>
sys@ORCL> alter session set nls_date_format=’DD/MM/YYYY HH24:MI:SS’;

Session altered.

sys@ORCL>
sys@ORCL> select sysdate from dual;

SYSDATE
——————-
29/03/2008 18:12:43

sys@ORCL>
sys@ORCL> alter session set nls_date_format=’DD/MM/YYYY’
2 select sysdate from dual;
select sysdate from dual
*
ERROR at line 2:
ORA-00922: missing or invalid option

Actually SET SQLBLANKLINES ON would have had that effect, as the new line wouldn’t have terminated the SQL command. Perhaps I should recommend that as a ‘good practice’ for SQL scripts.

This was not an interactive SQL*Plus session…it is in a .sql
script…so, I never hit enter at all

you would expect that sqlplus would keep asking for the semicolon. Instead it ignores (silently) the command when you just hit enter again.

[I remember that the 11i upgrade scripts had a
few bugslogged for queries on this table
during the upgrade process.]

In my SA/SE career I’ve had much greater success with many smaller systems making up a cluster (non-RAC, OS-level clusters), not only for greater overall uptime, but for the “invisibility-factor” of one node dropping out if there’s trouble or for maintenance. But the way our people want to use RAC here, the focus seems to be more on the HA benefits than HPC. So if we’re only going to be rolling out two nodes for HA, it seems like it would be better to fatten them up.

Do you intend to condense down? That is, do you intend to reduce the degree of horizontal scalability or will you tend to stay with the node count you are at and “fatten” the nodes?

I suppose if there is another tool that provides a more needed level of detail beyond what collectl can provide, and yes I realize such tools do exist 9-), perhaps the answer is to run both and use collectl to show what’s happening on the rest of the system during a test run assuming of course that the other tool provides timestamped history.

-mark

Cool. I’ve never “talked” to anyone that has these 4600s in production. I always presumed they’d work just fine–depending on workload of course.

CPU’s of the first ones purchased.

processor : 15
vendor_id : AuthenticAMD
cpu family : 15
model : 65
model name : Dual-Core AMD Opteron(tm) Processor 8218
stepping : 2
cpu MHz : 2600.027
cache size : 1024 KB
physical id : 7
siblings : 2
core id : 1
cpu cores : 2
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy
bogomips : 5199.32
TLB size : 1088 4K pages
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management: ts fid vid ttp tm stc

CPU’s of the second servers purchased:

processor : 15
vendor_id : AuthenticAMD
cpu family : 15
model : 65
model name : Dual-Core AMD Opteron(tm) Processor 8220
stepping : 3
cpu MHz : 2800.053
cache size : 1024 KB
physical id : 7
siblings : 2
core id : 1
cpu cores : 2
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy
bogomips : 5599.25
TLB size : 1088 4K pages
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management: ts fid vid ttp tm stc

16:43:20 up 288 days, 16:46, 3 users, load average: 11.49, 10.45, 8.63

The load average does not even give a taste of how well they run during our peak times (load over 30 for hours at a time). Glad to hear of others having the same success we are with this platform.

But, I’m a gadfly…oh, hold it, what about all those Google operations getting set up in the Columbia Gorge of the Pacific Northwest (US). Maybe they did some arithmetic

…thanks for stopping by Alex…as always.

Kevin, do you think the power and other direct costs are the most important? What about “let do the job to ones that can do it best”? I think that’s often forgotten with outsourcing in favor of cheaper labor, cheaper electricity and other direct cost factors.

Unfortunately, some (should I say most?) outsourcers (i.e. “insourcers” forget about it as well.

Btw, thanks for bringing our attention to those new AMD based rockets. I had a chance to play with 8 socket Barcelona modules half a year ago but those were terribly expensive and from another league I would say. DL 785 looks very attractive!

OCFS2, please no…
The only reason for OCFS would be shared export/import location or external tables and alike. Shared Oracle home would be another reason but it’s hardly a good idea anyway.

OCFS basically runs another clusterware behind the scene with it’s own heartbeat and eviction policy. Many many times I saw OCFS messing up CRS cluster and customer spending weeks to figure that out.

I would rather consider NFS or change design to be able to use local filesystem.

Re: raw devices. With 10gR2 you need to configure raw devices only during installation of 10.2.0.1 CRS - voting disks and OCR. After upgrade to 10.2.0.3, I always get rid of raw devices and use block devices directly. They are opened with O_DIRECT. I like to exclude unnecessary layers and raw devices is a good example.

I’m bit late on this but practical experience and all logic suggests that cheapest HA solution for Oracle is *single* instance. Adding physical standby to the picture improves availability a lot. Though, you won’t be able to run managed standby with SE, of course.

There is no way to have cheap HA with RAC but people threat term HA differently so you can say it depends.

I have been hired by Teradata just a week ago, while most of my professional life has been attached to Oracle in one or other way.

I swear I have nothing to do with this patnership…

Cheers.

Carlos.

a very important factor that is often forgotten in all the rationalizations about outsourcing the kitchen sink…

Excellent blog, Kevin. As usual.

I notice there was no mention of twitter or friendfeed, which are surely the only reason to require a data center these days…

Cheers

Tim…