Re: rman ? nfs ? slow

From: Ronald Rood <devnull_at_ronr.nl>
Date: Fri, 2 Jan 2009 20:26:09 +0100
Message-ID: <0001HW.C58428E1006D0B33F01846D8_at_news.individual.net>

On Fri, 2 Jan 2009 02:19:59 +0100, lfree wrote (in article <gjjq40$gus$1_at_news.cn99.com>):

>> I use rman backup product database on OS windows(2003 R2 standard 64
>> edition ) ,I check to find
>> it is slow (vs Linux rhel 4.3 ). My nfs server use SFU for windows.
>>
>> windows:
>>

>>> select * from v$version ;

>> BANNER
>> ----------------------------------------------------------------
>> Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
>> PL/SQL Release 10.2.0.4.0 - Production
>> CORE 10.2.0.4.0 Production
>> TNS for 64-bit Windows: Version 10.2.0.4.0 - Production
>> NLSRTL Version 10.2.0.4.0 - Production
>>
>> SELECT start_time, end_time, status, input_type, elapsed_seconds,
>> compression_ratio, TRIM (input_bytes_display) inputb,
>> TRIM (output_bytes_display) outputb,
>> TRIM (input_bytes_per_sec_display) input,
>> TRIM (output_bytes_per_sec_display) output,
>> TRIM (time_taken_display) TIME
>> FROM v$rman_backup_job_details
>> WHERE start_time > TRUNC (SYSDATE) - 1;
>>
>> START_TIME END_TIME STATUS INPUT_TYPE
>> ELAPSED_SECONDS COMPRESSION_RATIO INPUTB OUTPUTB INPUT OUTPUT TIME
>> ------------------- ------------------- ---------------- ---------- ------

> --


>> ------- ----------------- -------- -------- -------- -------- ----------

>> 2008-12-31 20:03:06 2008-12-31 20:19:37 COMPLETED        ARCHIVELOG

>> 991           2.59272 1.02G    402.88M  1.05M    416.29K  00:16:31

>>
>> output_bytes_per_sec_display = 416.29K.
>>

> 
>>> Hi,
> 
>>> I see you are using compression.  Are the CPU's saturated on the
>>> Microsoft server?
> 
> YES. I use compression.
> CPU is Intel Xenon X5460 3.16Ghz ,memory=8G.

Do you have resource manager enabled ?
I am working on an issue with a slow backup that happens when I have some other plan then the internal_plan active. If you ask me, the cpu - or core - count that resource manager is using is not the count we have in the system. With a custom plan active we get 3MB/s, with the internal_plan we get 34MB/s both cases are using 16 channel compressed backup sets to disk. The custom plan does give unlimited cpu to the sys_group at level 1.

In my case running linux x86_64 (8 dual core Sun x4600) with 10.2.0.4 database.

If this is also your problem you might see a lot of scheduler wait events for cpu quantum, while having plenty idle cpu left.

--
With kind regards / met vriendelijke groeten,
Ronald

http://ronr.blogspot.com/
http://ronr.nl/unix-dba

Received on Fri Jan 02 2009 - 13:26:09 CST