Pros & Cons of Using Raw Disk vs Cooked Files
Would anyone know of a good Tech Paper or Reference outlining the Pros and Cons of using raw disk devices as opposed to cooked files for IDS data storage on Unix systems? I am seeking some expert opinions on this subject as in my experience with Informix, I have never used cooked files. I need to be able to make a convincing arguement to our Systems Admin & Mgt why we choose the method we do (using raw devices). It has always been 'understood' that performance is supposed to be much better because the database server has direct I/O access to the devices, but I do not have a comparison. There is even a page or so in the IDS Administrators Guide which notes this and it also seems it is the IBM Informix recommmended practice. The issue is this: We have recently moved a major system from an HP-UX platform (IDS 7.31, 32 bit, XP256/XP128 SANs) to Sun (IDS 10.0, 64bit, EVA8000 SAN)and we've experienced a few problems when some work is done on the SAN that causes a 'blip' in disk availability (adding disk and a firmware upgrade, for example). It seems the database server does not handle a change of path for a disk (to OS disk mirror, I believe) and it loses its ability to read/access the chunks (I get the I/O read chunk errors in the online log and eventually when it gets to a critical chunk, On-Line aborts). Something like this may have happened once or twice over several years on the old system but has occurred four times now within two months on the new system. We also use Veritas. Systems Administration has suggested one alternative to preclude this issue from happening would be to convert Informix to cooked file systems over raw disk as they state a similar issue has never occurred on any of the devices or operating systems with applications or databases that use file systems. The other is to always take down the Informix instances before performing any SAN work (which could be inconvenient for business). I would like to be able to explain the advantages/disadvantages and need some suggestions as where to research further, as well as your experienced opinions. Thank You In Advance
We had somewhat similar problem, in our case it turned out to be a SAN issue, where the firmware at the SAN's end had to be upgraded. ----- Original Message ---- From: CYNTHIA SCOTT <clscott@klcorp.com> To: ids@iiug.org Sent: Tuesday, January 23, 2007 1:46:45 PM Subject: Pros & Cons of Using Raw Disk vs Cooked Files [8290] Would anyone know of a good Tech Paper or Reference outlining the Pros and Cons of using raw disk devices as opposed to cooked files for IDS data storage on Unix systems? I am seeking some expert opinions on this subject as in my experience with Informix, I have never used cooked files. I need to be able to make a convincing arguement to our Systems Admin & Mgt why we choose the method we do (using raw devices). It has always been 'understood' that performance is supposed to be much better because the database server has direct I/O access to the devices, but I do not have a comparison. There is even a page or so in the IDS Administrators Guide which notes this and it also seems it is the IBM Informix recommmended practice. The issue is this: We have recently moved a major system from an HP-UX platform (IDS 7.31, 32 bit, XP256/XP128 SANs) to Sun (IDS 10.0, 64bit, EVA8000 SAN)and we've experienced a few problems when some work is done on the SAN that causes a 'blip' in disk availability (adding disk and a firmware upgrade, for example). It seems the database server does not handle a change of path for a disk (to OS disk mirror, I believe) and it loses its ability to read/access the chunks (I get the I/O read chunk errors in the online log and eventually when it gets to a critical chunk, On-Line aborts). Something like this may have happened once or twice over several years on the old system but has occurred four times now within two months on the new system. We also use Veritas. Systems Administration has suggested one alternative to preclude this issue from happening would be to convert Informix to cooked file systems over raw disk as they state a similar issue has never occurred on any of the devices or operating systems with applications or databases that use file systems. The other is to always take down the Informix instances before performing any SAN work (which could be inconvenient for business). I would like to be able to explain the advantages/disadvantages and need some suggestions as where to research further, as well as your experienced opinions. Thank You In Advance ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum. ________________________________________________________________________________ ____ Cheap talk? Check out Yahoo! Messenger's low PC-to-Phone call rates. http://voice.yahoo.com