JaniceHetrick

20 Posts

1091

July 9th, 2009 16:00

Rebuild Rates on SATA II 1 TB Drives

I currently have 3 RAID 5 Groups 5 disks each of the 1 TB SATA II Drives...has anyone had any rebuild experience on these or something close? I have a CX3-40c.

Responses(8)

AranH1

2.2K Posts

0

July 13th, 2009 15:00

First, I would highly recommend that you have a hot spare for every type of drive in your array. Deploying SATAII drives without a hot spare is not a good idea in my opinion and would leave a RAID group in a degraded state until you receive a replacement drive.

Second, I think your fear that a SATAII hot spare will replace a failed FC drive is unfounded. Yes, technically a SATA hot spare can replace a failed FC drive, but as long as you have sufficient FC hot spares in the array of the appropriate size then a SATAII hot spare will not be invoked for a FC drive. In the past four years I have been managing multiple CLARiiONs with FC and SATAII drives and have never had this happen. Plan your hot spare assignment according to EMC best practices, which translates to one hot spare for every 30 drives.

Regarding your RG layout a third option would be two RAID5(6+1) and one hot spare. That would spread your LUNs over more spindles and reduce the amount of overhead lost to RAID parity for that enclosure.

dynamox

1 Rookie

•

20.4K Posts

1

July 9th, 2009 17:00

search powerlink for document "The influence of Priorities on EMC Clariion LUN Management Operations"

RyanP2

261 Posts

0

July 10th, 2009 07:00

The doc is actually located under the white papers for the CX3 but I can't find it under CX4 which I think it should be there too.

Not taking credit for the answer but if your having problems locating the doc here is the link:

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H4153-the-influence-of-priorities-on-lun-mgmt-optns-wp.pdf

-Ryan

ZaphodB

195 Posts

0

July 10th, 2009 10:00

A correct answer would be that "it depends", and a good thing to understand would be some of the things which it depends on.

Here are a few things that come to mind:

A rebuild will only be done if a drive fails unexpectedly. I know that sounds obvious, but routinely a drive may invoke and be copied to a hot spare because of error threshholds and not because it has actually failed. A copy is significantly faster than a RAID5 rebuild.

Whether it is copied or rebuilt, only bound (allocated) sections of the drive will be created. So if you have a RAID group which is 100% in use then the whole drive will be re-created, but if you have unbound space in the RAID group then the amount of data copied will be less. Note that the unit does not know if the bound space contains anything, but it does know whether it is part of a LUN or not.

Rebuild rate is a characteristic of the RAID group. You can dictate that important groups rebuild at a faster rate than less important groups.

Rebuilds are a background task for the unit, so how busy the Clariion factors in, as well as how busy the RAID group is. If the Clariion is relatively busy the rebuild may take longer. If the RAID group is busy, the rebuild may actually move a bit quicker if the background process is opportunistic in rebuilding what is required for ongoing I/O requests.

AranH1

2.2K Posts

0

July 13th, 2009 12:00

What is it that you are concerned about, rebuild times?

JaniceHetrick

20 Posts

0

July 13th, 2009 12:00

So, thus far, is there anyone in the forums that has implemented a shelf of SATA II 1 TB drives with 3 sets of RAID 5 and has had to rebuild a RAID set?

I understand there are many factors to rebuilds

JaniceHetrick

20 Posts

0

July 13th, 2009 14:00

Yes and best practices here is my scenarios:

Scenario 1:
One shelf of 15 1TB SATA II drives that I plan on setting up as such:
Disks 0-4 RAID 5 (4+1)
Disks 5-9 RAID 5 (4+1)
Disks 10-13 RAID 5 (3+1)
Disk 14 Hot Spare

OR

Scenario 2:
One shelf of 15 1TB SATA II drives that I plan on setting up as such:
Disks 0-4 RAID 5 (4+1)
Disks 5-9 RAID 5 (4+1)
Disks 10-14 RAID 5 (4+1)

Issue...Scenario 1 with the hot spare I have found maybe an issue since a SATA II drive could be used for one of my FC drives (for which I have 2 hot spares already)

Issue...Scenario 2 without a hot spare I need to know how long the rebuild rate will be on a 4+1 RAID 5 (thanks to a previous post I have that calculation)

Main concern is do I spread the RAID groups throughout the shelf or keep them as I listed above (0-4, 5-9, 10-14) or from reading through best practices it seems that the RAID groups are spread out within the shelf.

Really what I am trying to do is get the best bang for my buck and keep the rebuild within reason. RAID 6 is out for now due to I have FLARE 24 not 26 on my CX3-40c.

Any input is appreciated!

kelleg

4.5K Posts

1

July 15th, 2009 07:00

Another couple of documents that are very helpful:

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

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

Also, if you feel that your questions have been answered could you please mark the question as answered, awarding points to the replies that best answered your question.

glen

View All

No Events found!