It is recommended to segregate the drive into multiple pools based on
- workload - sequential workload from random workload, large block size from small block size; LUNs that require FAST cache or FASTVP from LUNs that do not require (sometime for cost reason)
- availability - LUNs that require higher RAID protection (higher cost) from lower RAID protection (lower cost)
Please consider moving this question to the proper forum for maximum visibility. Questions written to the users' own "Discussions" space don't get the same amount of attention and questions can go unanswered for a long time (if at all). You can do so by selecting "Move" under ACTIONS along the right. Then search for and select: "VNX Support Forum". There is also a recent post demonstrating the steps to do the same:
I see that you had found the answer on your own and have written up a nice summary. Not to take anything away from what you have, but I'd also like to suggest the following document: "EMC VNX Unified Best Practices for Performance Applied Best Practices Guide".
Skip to the section "Storage pool considerations" > "Storage pool creation"
1) The maximums are exactly the supported maximum drive count - 4 (vault)
2) It also shows the maximums per model you can add at a time
3) It then discusses why you would want separate pools:
a) Fault domains (proverbial "all your eggs in one basket")
b) Separate workloads
Now what isn't documented specifically in this PDF are some specific considerations (but covered under the general considerations above):
1) Separating file storage from block (dedicated pool for LUNs presented to ~filestorage) per the calculation under the section "Considerations for VNX for File"
2) Remember FAST Cache is enabled/disabled at the pool level and not individual LUN level (except for traditional FLARE LUNs), so important to keep that in mind
- for instance, small block sequential workloads aren't generally the best candidates (FLARE 32 there is the avoidance algorithm but being reconsidered for VDI workloads - patch available)
3) Separating backup LUNs from production LUNs
4) Contention - for example, separating Reserved LUNs (collectively RLP) if decision made to use DLU (thick) pool LUNs and not traditional RG LUNs (FAST VP disabled of course) from the production LUNs they are associated with
Then you get into the conversation about traditional RG's vs Pools.
engkong
1 Rookie
•
11 Posts
0
June 30th, 2013 02:00
It is recommended to segregate the drive into multiple pools based on
- workload - sequential workload from random workload, large block size from small block size; LUNs that require FAST cache or FASTVP from LUNs that do not require (sometime for cost reason)
- availability - LUNs that require higher RAID protection (higher cost) from lower RAID protection (lower cost)
christopher_ime
4 Operator
•
2K Posts
0
July 4th, 2013 00:00
engkong,
Please consider moving this question to the proper forum for maximum visibility. Questions written to the users' own "Discussions" space don't get the same amount of attention and questions can go unanswered for a long time (if at all). You can do so by selecting "Move" under ACTIONS along the right. Then search for and select: "VNX Support Forum". There is also a recent post demonstrating the steps to do the same:
EMC Community Network Frequently Asked Questions
I see that you had found the answer on your own and have written up a nice summary. Not to take anything away from what you have, but I'd also like to suggest the following document: "EMC VNX Unified Best Practices for Performance Applied Best Practices Guide".
https://support.emc.com/docu42660_Applied_Best_Practices_Guide:_EMC_VNX_Unified_Best_Practices_for_Performance.pdf
Skip to the section "Storage pool considerations" > "Storage pool creation"
1) The maximums are exactly the supported maximum drive count - 4 (vault)
2) It also shows the maximums per model you can add at a time
3) It then discusses why you would want separate pools:
a) Fault domains (proverbial "all your eggs in one basket")
b) Separate workloads
Now what isn't documented specifically in this PDF are some specific considerations (but covered under the general considerations above):
1) Separating file storage from block (dedicated pool for LUNs presented to ~filestorage) per the calculation under the section "Considerations for VNX for File"
2) Remember FAST Cache is enabled/disabled at the pool level and not individual LUN level (except for traditional FLARE LUNs), so important to keep that in mind
- for instance, small block sequential workloads aren't generally the best candidates (FLARE 32 there is the avoidance algorithm but being reconsidered for VDI workloads - patch available)
3) Separating backup LUNs from production LUNs
4) Contention - for example, separating Reserved LUNs (collectively RLP) if decision made to use DLU (thick) pool LUNs and not traditional RG LUNs (FAST VP disabled of course) from the production LUNs they are associated with
Then you get into the conversation about traditional RG's vs Pools.