The test platform:
The details of the components of the test could go on forever, so I'll just give a quick overview. The IBM P5 595 has 8 online CPU's, with 32GB of ram on each. I had 4 HBA's connecting it to the USP-V through 2 VIO servers and a director/switch. There are no other lpars running on this frame or using this storage.
We'll be testing multiple tiers for storage, including:
1. EMC Clariion 10k FC Raid 5
2. HDS AMS with 7.5k Sata Raid 5
3. USP-V internal 10k FC Raid 5
4. USP-V internal 10k FC Raid 10
5. USP-V internal SSD Raid 5
The USP-V has 512G of configurable cache that can be used as a read, write or read/write cache. It has internal storage capacity of 332TB, and the ability to virtualize other storage units "behind the cache", so we're able to have the Clariion unit serve storage-cached, virtualized and thin provisioned, behind the USP-V. In addition, it has the ability to perform ShadowImage copys for clones/db refreshes and syncs. It has the ability to find hot luns and move them from tier to tier, based on a script. In theory, it should bring a near-zero MTTR for backups, it should increase availability of non-prod environments with instant refreshes, and it should allow us to store data much more densely, effectively reducing our projected storage growth rate (and expense). With the additional cache, it should increase performance for our database servers.
Potentially, it will bring all these additional features and provide storage cost avoidance that will more than pay for itself...but if it doesn't perform at least as fast as the Clariions, its out the door, so let's take a look at the performance first.
This will be a 3 part series covering performance, general features and thin provisioning:
In this series:
HDS USP-V Overview for Oracle DB
HDS USP-V Performance for Oracle DB
HDS USP-V Features for Oracle DB
HDS USP-V Thin Provisioning for Oracle DB