- Community Home
- >
- Servers and Operating Systems
- >
- Operating Systems
- >
- Operating System - HP-UX
- >
- Effect of disk_asynch_io=true on Oracle Instances ...
Categories
Company
Local Language
Forums
Discussions
Forums
- Data Protection and Retention
- Entry Storage Systems
- Legacy
- Midrange and Enterprise Storage
- Storage Networking
- HPE Nimble Storage
Discussions
Discussions
Discussions
Forums
Forums
Discussions
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
- BladeSystem Infrastructure and Application Solutions
- Appliance Servers
- Alpha Servers
- BackOffice Products
- Internet Products
- HPE 9000 and HPE e3000 Servers
- Networking
- Netservers
- Secure OS Software for Linux
- Server Management (Insight Manager 7)
- Windows Server 2003
- Operating System - Tru64 Unix
- ProLiant Deployment and Provisioning
- Linux-Based Community / Regional
- Microsoft System Center Integration
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Community
Resources
Forums
Blogs
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 05:17 AM
тАО03-15-2005 05:17 AM
Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 05:20 AM
тАО03-15-2005 05:20 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 05:23 AM
тАО03-15-2005 05:23 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
Is this true?
I quote a response:
"The value below is the default and it means oracle will try to use asynchronous IO if possible. With filesystem oracle uses threaded asynchronous IO only when the operating system supports it. But on filesystems we can simulate threaded asynchronous IO in Oracle by using more DBWn writers"
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 06:25 AM
тАО03-15-2005 06:25 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
If you need more I/O however, there are things you can do. Definitely you can support more than one DBWriter - no matter how small your system is. I generally allocate one or so per filesystem - but try not to exceed number of procs. But, the real answer is to start up the statspack, get some baseline, then start changing the number of db writers you have and see if the delays from writes start coming down. Your ability to support a 'bunch' of db writers really depends on how much bandwidth you've got to the storage system (in terms of concurrency). Can your storage system support lots of concurrent writes ? Then it's time to start increasing and measuring.
Another thing I like to do to increase my ability to write thing concurrently is to pre-allocate my tablespaces across many files, so that as data grows - it grows across the files in the tablespace - giving me some approximation of concurrency - at times. So, if you need a new file added to your tablespace (and say it's a hot one), don't just add a 2G file - why not add 4 512MB files or even 8 256M B files instead (Check that your nfile parameter can handle it)? Even better, if this tablespace is really hit hard - why not put those files across several mount points and interleave them? This will give you a lot of disk mechs doing your bidding for you. You can watch as the data grows over time in the files, it will be pretty evenly spread.
Keep in mind the load your storage system can handle - on direct attached I/O with ultra-SCSI I'll often use a db_writer_processes value of 4 or 6 (depending on how many disks/controllers/cache I'm using), but on a system with large enterprise storage - I'll use a value of 16 or 24.
As long as my total I/O doesn't go red-line on me, and my statspack shows better response times on delivering I/O and no ugly adverse effects in the rest of the system - I'll keep pushing db_writers until I'm happy with it.
Also, you can use one DB_Writer with a bunch of I/O slaves. I've not seen when using synchronous I/O myself on how that's better, or worse for that matter. In my researching papers, etc on the topic, I've seen how I/O slaves is MUCH better when doing asynchronous I/O in raw partitions. Keep in mind that when using multiple I/O slaves - you'll only have one DB writer ... hmmm ... is that what you have now??? If so, this is all a non-discussion...
You could do either - increase db-writers or increase with I/O slaves - either one will get you there. But, measure and test your way (or at least track) through it. Statspack reporting will help you figure it out.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 08:12 AM
тАО03-15-2005 08:12 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
Had to set up a file called /etc/privgroup
That let let oracle run in its supposed asynch mode.
I don't know whether its getting asynch i/o or not, but the documenation seems to be trying to tell us that it is.
Its probably some form of psuedo asynch or they devide up the blocks and let multiple processes work on the same tablespace at the same time.
Your dba is probably a little confused. As we've discussed going raw i/o is not going to provide enough performance boost to justify your time.
SEP
Owner of ISN Corporation
http://isnamerica.com
http://hpuxconsulting.com
Sponsor: http://hpux.ws
Twitter: http://twitter.com/hpuxlinux
Founder http://newdatacloud.com
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 10:32 PM
тАО03-15-2005 10:32 PM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
as far as know you must be on raw device to support async IO (That's Oracle prospective).
I async IO exists on fs, I don't know Oracle's position (support, ...)
check this thread
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=753721
Also, you may want to check DBWR_IO_SLAVES parameter as an alternative.
check Metalink doc id 97291.1 "DB_WRITER_PROCESSES or DBWR_IO_SLAVES?"
Regards
Jean-Luc
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО03-15-2005 11:16 PM
тАО03-15-2005 11:16 PM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
It is quite simple:
You can set disk_asynch_io to TRUE (I've it setted) but it won't work unless you REALLY HAVE RAW DEVICES and all the HP-UX stuffs right configured (see Metalink note 139272.1)!
If you don't have raw devices, Oracle will do synchronous I/O EVEN WITH disk_asynch_io = TRUE!
Best Regards,
Eric Antunes
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО05-02-2005 02:03 AM
тАО05-02-2005 02:03 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
There is one thing I've discovered last week. If your OS don't support asynchronous I/O it is better for performance to set disk_asynch_io = false (I will see this at checkpoints). With disk_asynch_io = true you will have much more CPU work!!
Best Regards,
Eric Antunes
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО05-02-2005 02:06 AM
тАО05-02-2005 02:06 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
Thanks.
But did you verify this from Oracle that this indeed is the case? Our DBAs are all in concert in saying it will just be ignored since it is not active and we're running on cooked storage.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
тАО05-02-2005 02:15 AM
тАО05-02-2005 02:15 AM
Re: Effect of disk_asynch_io=true on Oracle Instances that are not on RAW Storage
I've found note 1016262.102 (Excessive CPU Usage by Checkpoint Process) when I was searching for I/O issues...
Here is part of the Note:
...
Explanation: ============ The "init
...
I've changed it and now I've less CPU consumption at checkpoints!!
Eric