For best performance keep raid groups small (4 or 5 disks) and create metaluns with components in each raid group. With 14 disks I would create 2 x (6+1) and for each required LUN I'd create 2 LUN's each of which is half the size of what the host needs and create a striped META of those.
The benefit of having smaller RG's is that a defective disk will only have an impact on that RG and a rebuild is much faster on a small RG since it only has to read the data from a few disks instead of 13 ones or so.
Consider this:
A (4+1) RG has room for data on 4 disks, but the performance of 5 disks
Having a striped META on 2 (4+1) RG's has room for data on 8 disks, but performance of 10 disks
1 (8+1) RG has room for data on 8 disks, but the performance of 9 disks, instead of 10.
As previously stated, the maximum number of drives configurable into a single RAID group is 16. However for availability reasons it would be imprudent to configure a 15+1.
The optimum maximum size of a RAID level-5 group is dependent on the I/O and heavily influenced by the block size. It is also important to consider the effect of a rebuild, and not make them too large.
If your I/O is large block sequential, or has a large sequential component the maximum would be an 8+1. This construct leverages the CLARiiON's attempts to perform full-stripe writes.
The optimum maximum RAID group size for random I/O is more problematic. How many IOPS do you need? A truely random small block workload would have no problem with a 15+1. However, a rebuild would take a very long time. (This is the availability concern mentioned above.) If had to give my opinion, for a truly random workload any RAID-5 configuration between and including 4+1 and 8+1 works (IOPS dependent). For example, a 6+1. However, if there is any sequential component (its really a mixed work load), the 8+1 is as large as I'd go. This is purely to keep the rebuild times low.
Information on RAID group sizing is found in the RAID groups and I/O block size section of the EMC CLARiiON Best Practices for Performance and Availability, FLARE Rev. 29. In addition, there is a further discussion of RAID group size in the CLARiiON RAID groups section of the EMC CLARiiON Storage System Fundamentals for Performance and Availability document. These documents are available on Powerlink.
Thank You for the answer. I have only 15 disks left in the array if i go for 2(6+1) disks i will be left with one disk doing nothing. Can you suggest me more best option. Thanks.
It is not a Hot spare, but i have disks across different DAE's. I will try 1*6+1 and 1*7+1 and see the performance on the servers as these are test servers. Thank you for your time.
Do you know what the workload that you will be putting on the new LUNs? It's best to determine what the total workload will be before you configure the raid groups and LUNs - see the Best Practices guides for information on sizing the storage requirements based on workload.
Just be sure if you configure two different sized RGs that you do not use them to build Metas as previously recommended. The RGs underlying Metas should be the same size and configuration (again as per Best Practices documents)
If you want to be able to build the Metas then you should probably take the cue from RRR and configure 2 x RAID5(6+1) RGs and configure the last drive as an extra Global Hot Spare. Unless you have WAY to many already it can never hurt to have one more :-)
dynamox
9 Legend
•
20.4K Posts
0
April 7th, 2010 12:00
depends on what you are trying to do, take a look at Performance and Availability documents here
https://community.emc.com/community/support/clariion?view=documents
bonnu06
2 Intern
•
258 Posts
0
April 7th, 2010 12:00
Thank You
bonnu06
2 Intern
•
258 Posts
0
April 7th, 2010 12:00
Can i use 7+1 disks for Raid 5 or 14+1 for better performance.
dynamox
9 Legend
•
20.4K Posts
0
April 7th, 2010 12:00
16 drives per raid group
RRR
4 Operator
•
5.7K Posts
0
April 8th, 2010 01:00
For best performance keep raid groups small (4 or 5 disks) and create metaluns with components in each raid group. With 14 disks I would create 2 x (6+1) and for each required LUN I'd create 2 LUN's each of which is half the size of what the host needs and create a striped META of those.
The benefit of having smaller RG's is that a defective disk will only have an impact on that RG and a rebuild is much faster on a small RG since it only has to read the data from a few disks instead of 13 ones or so.
Consider this:
A (4+1) RG has room for data on 4 disks, but the performance of 5 disks
Having a striped META on 2 (4+1) RG's has room for data on 8 disks, but performance of 10 disks
1 (8+1) RG has room for data on 8 disks, but the performance of 9 disks, instead of 10.
jps00
2 Intern
•
392 Posts
0
April 8th, 2010 05:00
As previously stated, the maximum number of drives configurable into a single RAID group is 16. However for availability reasons it would be imprudent to configure a 15+1.
The optimum maximum size of a RAID level-5 group is dependent on the I/O and heavily influenced by the block size. It is also important to consider the effect of a rebuild, and not make them too large.
If your I/O is large block sequential, or has a large sequential component the maximum would be an 8+1. This construct leverages the CLARiiON's attempts to perform full-stripe writes.
The optimum maximum RAID group size for random I/O is more problematic. How many IOPS do you need? A truely random small block workload would have no problem with a 15+1. However, a rebuild would take a very long time. (This is the availability concern mentioned above.) If had to give my opinion, for a truly random workload any RAID-5 configuration between and including 4+1 and 8+1 works (IOPS dependent). For example, a 6+1. However, if there is any sequential component (its really a mixed work load), the 8+1 is as large as I'd go. This is purely to keep the rebuild times low.
Information on RAID group sizing is found in the RAID groups and I/O block size section of the EMC CLARiiON Best Practices for Performance and Availability, FLARE Rev. 29. In addition, there is a further discussion of RAID group size in the CLARiiON RAID groups section of the EMC CLARiiON Storage System Fundamentals for Performance and Availability document. These documents are available on Powerlink.
bonnu06
2 Intern
•
258 Posts
0
April 8th, 2010 05:00
Hello,
Thank You for the answer. I have only 15 disks left in the array if i go for 2(6+1) disks i will be left with one disk doing nothing. Can you suggest me more best option. Thanks.
bonnu06
2 Intern
•
258 Posts
0
April 8th, 2010 06:00
It is not a Hot spare, but i have disks across different DAE's. I will try 1*6+1 and 1*7+1 and see the performance on the servers as these are test servers. Thank you for your time.
Regards
Bonnu
RRR
4 Operator
•
5.7K Posts
1
April 8th, 2010 06:00
1 hot spare ? Or 1 x (61) plus 1 x (71)
kelleg
4 Operator
•
4.5K Posts
0
April 8th, 2010 09:00
Do you know what the workload that you will be putting on the new LUNs? It's best to determine what the total workload will be before you configure the raid groups and LUNs - see the Best Practices guides for information on sizing the storage requirements based on workload.
glen
jps00
2 Intern
•
392 Posts
0
April 8th, 2010 10:00
^^^ ++
Allen Ward
4 Operator
•
2.1K Posts
0
April 8th, 2010 12:00
Just be sure if you configure two different sized RGs that you do not use them to build Metas as previously recommended. The RGs underlying Metas should be the same size and configuration (again as per Best Practices documents)
If you want to be able to build the Metas then you should probably take the cue from RRR and configure 2 x RAID5(6+1) RGs and configure the last drive as an extra Global Hot Spare. Unless you have WAY to many already it can never hurt to have one more :-)