DMX - AIX & Oracle throughput MB/s

RAID-S? Haven't heard about that for some time..

What are they doing when they measure 10MB/sec? Backup? Queries?

Performance is determined by the underlying setup of the disk itself. If this is a database with heavy writes, and the underlying disk structure on the EMC is raid 5, then you will encounter performance problems whenever writes get heavy.

If you can afford it, try and get data/index/redo logs onto raid 10 configured EMC disk, it this is an (OLTP) system.

And again this document may be helpful for you.

http://www.emc.com/collateral/hardware/solution-overview/h2603-oracle-db-emc-symmetrix-stor-sys-wp-ldv.pdf

Dont forget to assign points and mark it as answered if you get the solution!

And RAID-S or Parity RAID was introdouced before Raid-5 parity raid was implemented in lowcost environments.. Parity information is written on a single hyper volume. and it is configured in 3+1 (75 percent usable) and 7+1 (85 percent usable) Parity Raid is alien to current DMX configurations.. instead Raid-5 is used.

Regards

A small correction, Raid-S was used in Symmetrixes, as EMC didn't have Raid-5 until later... Ie. Symm 8000 -series only had Raid-S.

A small correction ..

You can find boxes running 5670 (probably close to EOSL) that have both RAID-S and RAID-5 volumes

I guess EMC offered RAID-S with older models since it was the best compromise between protection/waste of space/performances. With recent models (different CPU, different backend technology, different everything) EMC offered also RAID-5. Now we are offering even RAID6...

EMC is constantly changing the offering.. The reasons for that changes may be difficult to understand.. But there are reasons

Reason 1: competetion
Reason 2: competetion
....

As I said .. reasons

The initial throughput they were seeing was when they were doing a database load. Since then has dramatically increased, over 100 MBs. I am not sure what took they were utilizing to measure the througput. At the time our DMX was installed RAID-5 was not supported on the box. We are trying to "make do" with our current RAID-S configuration until we have a technology refresh so changes to the DMX side of the house is not an option at this point. When allocating hypers to the AIX hosts we have taken care to make sure back-end distribution is optimal. What I'm looking for at this point is more AIX and Oracle tuning information that may be helpful to the DBA and AIX admins.

Thanks...

RAID3 typically stripes data for a single logical volume across all drives in the raid group, and puts the parity on one drive. EMC RAID-S puts all the data for one logical on one drive, and the parity for multiple logical volumes in a raid group on one drive.

I always wondered why EMC didn't call their RAID-S simply RAID-3.....

Unfortunately we don't have the spare capacity to reconfigure to RAID-5 so looking for improvement without reconfiguration.

RAID-S and good write performance don't go together. RAID5 could improve this significantly because of optimized writes for sequential IO. All DMX systems can support RAID-5 with 5670 code or above.

I never quite understood how RAID-S worked.... I'm glad EMC now uses the RAID standards

I love trainings..... Haven't had any in 2 years or so.... but then again.... RAID-S is ancient history. I know they tried to teach me about it years ago, but you simply forget if you're not using it

Rob the good old RAID-S is quite easy to understand if you have a good teacher .. Don't you want a training on our RAID-S ??

Oh BTW there is still someone from Kiev waiting for a training on drives, cyls, heads .. thus you have to wait a little