cx3-20f RAID creation best practice

Question

experts,

we have clarioon cx3-20f disk array with two DAE.

DAE 1 has 15 FC disks ( 268GB)

DAE 2 has 15 SATAII disks (458GB)

Luns created from disarrays needs to be presented to our RAC database servers (2 physical servers) and vmware ESX 5.0.

i am planning to setup RAID 5. Please let me know best practice for creating RAID 5 out of this two enclosure with 2 different types of disk drives.

i am planning for something like following. ( ALL RAID 5)

RAID Group 0 -->with 5 FC disk [ FC disk 0, FC disk 1, FC disk 2, FC disk 3, FC disk 4]

RAID Group 1 -->with 5 FC disk [ FC disk 5, FC disk 6, FC disk 7, FC disk 8, FC disk 9]

RAID Group 2 -->with 4 FC disk [ FC disk 10, FC disk 11, FC disk 12, FC disk 13]

RAID Group 50 --> with 1 FC disk . hot spare disk [ FC disk 14]

RAID Group 3 -->with 5 SATAII disk [ SATAII( disk 0, disk1, disk2, disk3, disk4)]

RAID Group 4 -->with 5 SATAII disk [SATAII ( disk5 , disk6, disk7, disk8, disk9)]

RAID Group 5 -->with 4 SATAII disk [ SATAII ( disk 10, disk11, disk12, disk 13)]

RAID Group 51 --> with 1 SATAII disk . hot spare disk [ SATAII disk 14]

Please let me know whether above setup looks good or advise otherwise with best practice.

Storagesavvy · Answer

If you are planning to use ASM for RAC, then you will probably want each LUN in the ASM Group to be carved from identically configured RAID Groups. That being the case, you will probably want to do something like two 61 groups instead of two 41 plus a 3+1.

What sort of LUN combinations are you looking to create? Do you need VMFS and Raw Devices? Just ASM LUNs? Non-ASM? Etc? We’ll need a bit more information about what you are trying to provision to make any sort of real recommendation.

Richard J Anderson

sansaran1 · Answer

Thanks rich for the reply

we will be having combination of ASM. ocfs2 and VMFS for ESXi5

Physical server lun assignment

we will be running Oracle RAC on 2 physical servers( 2 node cluster). for that we need ASM and ocfs2(requested by dba's) lun and that need to shred between 2 physical servers.

ESXi server lun assignment

we have 2 esxi host in cluster. not yet lun presented to esxi host. our plan is to have multiple VM guests. Also we need to setup ocfs2 cluster for DBAS shared between two VM's in the cluster. i am planning to use raw device mapping ( in physical mode) for ocfs2 lun.

Please povide recommendations.

Storagesavvy · Answer

So the database will be on ASM it sounds like.. Will there be VMs other than the Oracle servers? Will the VMFS only host the Oracle OS disks or are there other workloads? Any idea on what the performance requirements will be for the Physical and Virtual databases? Performance for the data that will live on the VMFS?

Richard J Anderson

sansaran1 · Answer

As of now , no plans for any other VM's other than oracle on ESXi cluster. so VMFS hosts only Oracle os disks. it is going to be our DR and lot of sync will happen from prod to DR through dataguard on physical database servers.

creating stripped metaluns across raid group will increase performance or it is a extra burden for storage processor.

Storagesavvy · Answer

This is a small enough configuration that MetaLUNs would not be a problem in my mind. Since this is a dataguard replication destination, I’d use the FC disks for the Database and the SATA for the OS disks.

A couple things…

1.) ASM stripes data across multiple LUNs so in that case you just create a LUN in each RAID Group and present to the RAC servers, then ASM will group them together. BE SURE to properly set the offset (64KB or 128 blocks) for the partition with fdisk, even if you are presenting it as a Raw Device to ASM. Misalignment causes extra IO and overhead. Don’t create more than 1 LUN per RAID group for the same ASM group, this will cause disk head thrashing and kill performance. If you have multiple ASM groups, it’s okay to provision a LUN for each ASM Group from each RAID Group.

2.) For the VMFS, you could create two RGs and then use MetaLUNs across those RGs to create the datastore, or create LUNs in each RG and have multiple datastores. Overall IOPS will be higher for a single datastore on a MetaLUN compared to one in an RG, but you could get better combined performance with multiple LUNs up to a point. SATA is not that great at multi-threaded IO.

It sounds as if you only have the two shelves of disk; if that’s the case then the first 5 x 300GB drives would also be the OS-Vault disks for the CX3-20 FLARE Code. That will decrease the available space on those disks as well as reduce overall IOPS available. You could put the Oracle VM Disks there, and then create a single 8+1 for the databases leaving the last disk for a spare. ASM uses large stripes for the DB files so performance on a larger RG should be okay. You can then provision the SATA disks as an Archive/Flashback area or use for VMFS.

If I recall correctly the OCFS volume is not super IO intensive and is pretty small, you should be able to find a home for it among the remaining space. A lot of it depends on how your DBA wants the ASM Groups laid out.

Richard J Anderson

kelleg · Answer

A couple of documents available on PowerLink would also be useful:

EMC CLARiiON Best Practices for Fibre Channel Storage: FLARE Release 26 Firmware Update - Best Practices Planning

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H2358_clariion_best_prac_fibre_chnl_wp_ldv.pdf

EMC CLARiiON Storage System Fundamentals for Performance and Availability

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H1049_emc_clariion_fibre_channel_storage_fundamentals_ldv.pdf

glen

CLARiiON

cx3-20f RAID creation best practice

Was this post helpful?