We have a K580 with an array 12H (autoraid) with 12 18Gb HDD?s On this array we have created an Oracle database with a total storage of 35 Gb. The database is spread over two LUNS of 23 Gb each with crossed primary / alternate SCSI controllers. The unallocated space would be about 14 Gb. After creating the database we saw that all the database files are written in RAID 5 and not in RAID 1. When we created it the RAID 0/1 space was a several 100 Mb?s. Above is a test situation for the porting of one of our applications to HP-UX and Oracle. When we go in production the unallocated space for the autraid would be ? 70 Gb. The question is now : is there a way to get de databasefiles default written in RAID 0/1 even as the RAID 0/1 space allocated is just a several 100 Mb?s and the write working set is in GB (logprint info) Please tell what i must do to get a good performance from the start.
The 12H will move files from Raid 5 to Raid 0/1 on the basis of them being requested. At the start (eq, restore) everything will be simply written to disk. When in use the 12H will re-allocate files and build up the Raid 0/1 space. When we reconfigured our 12H and restored the data the initial raid 0/1 space started at 9Mb (from a previous 25000Mb). This has increased to 22000Mb at present. I have included the script I use to monitor this build-up. This is what the 12H will do automatically and over which the user has no control. If you want the 12H to start out with a large Raid 0/1 space I suggest you give it something to do (start copying files, fill up the volumes, delete the files again) and you will see Raid 0/1 space grow automatically. Using Logprint you can monitor the Raid 5 to 1 reallocation info to see how it is doing. Other than this, I know of no other way to 'force the 12H' to go to a specific state. This, in fact, is also what the 12H was designed for, to take that job out of the hands of the administrator. Something, I must admit, is/was not always to my liking.
Hi In general the AutoRAID manages the 0/1 space for you. You do not have any control over it. In general the more LUNs the better. Here is some additional info:
Optimizing Performance Disk array performance is influenced by many factors, some host-related, some related to the configuration of the disk array itself. Host-related disk array performance issues can typically be identified and corrected using operating system utilities or third-party applications designed for this purpose. The following list identifies some of the factors that directly influence disk array performance. Other factors, such as a rebuild, can temporarily affect array performance. + Amount of available RAID 0/1 ? In most cases, the more RAID 0/1 space available, the better the disk array performance. With enough RAID 0/1 space, the disk array can service all write I/Os from RAID 0/1, thus allowing an optimum write working set. RAID 0/1 space can be increased by increasing the amount of unallocated capacity on the disk array. + Enabling Active Hot Spare ? With Active Hot Spare enabled, the disk array can recover quickly from a disk failure without the need for any action on your part. This maintains performance because while the disk array is operating in non-redundant or critical mode, performance decreases. Active Hot Spare provides an additional performance benefit because it is used as RAID 0/1 space until a disk fails. + Amount of cache DRAM ? Increasing the amount of cache DRAM can alleviate bottlenecks that may form in the controller data path. This increases the rate at which the disk array can service host I/Os. + Proper disk module installation ? Installing the disk modules in the correct slots in the enclosure distributes the disks across the internal SCSI busses and balances the I/O load properly. Disk modules should be installed from left-to-right, top-to-bottom. + Keep the disk capacity small ? Having twelve disks with smaller capacities will improve the performance as compared to having twelve disks with larger capacities.
If you are really keen on performance, Cees, you should get rid of 12h. Buy 2 enclosures and simply mirror your 12 disks for netto 12*9=108GB. The AutoRAID will always stand in the way of performance.
Unfortunately with the AutoRaid devices there is not a way to explicitly define what uses Raid 0/1 and what uses Raid 5. The AutoRaid has some 'smarts' built in that causes it to automatically move data from Raid 5 to 0/1 or vice versa as it sees fit.
I have two AutoRaid 12H's and I have never really been pleased with their performance. It is partially because of the Raid 0/1 and Raid 5 issue and partially because the SCSI bus to the system is FWD/DIFF and the internal SCSI bus is SE.
You can force the autoraid 12h to go into Raid 0/1 configuration by allocating the total lun space less or equal to half the space. In this case the autoraid will work exclusively in Raid 0/1. Hope this will help.
Indeed, since it is AUTOraid you do not have control about RAID levels (with an FC-60 or XP, you would have). If you need everything in RAID 0/1 than do not assign more than: 12*18GB minus 1* 18Gb (for hot spare) and take 50% of this= 99Gb to LUNS. Creating 2+ LUNS in the volumegroup and striping across the LUNS can speed up I/O's.
At this moment your database is in RAID 5, but the autoraid will balance itself (when there's none/few I/O traffic), your database might be in RAID 0/1 in a few days time. The autoraid will try to keep the least accessed data in RAID 5 to get full performance and the most frequent accessed data (which will be in 0/1).
