Hi All, I need some advise from admins with experiance .We have installed a new VNX5500 and are looking to provision storage pools with a mix of SAS - 300gb and 600gb 15K drives, NL SAS 2TB 7.2K.
I ‘ve-
92+4(vault) 300GB SAS
30 600GB SAS
9 2TB NL SAS
Customer has 4 new HPUX running oracle +custom application they need 300GB SAS RAID5 & 2TB RAID6 NL SAS configuration for this
And , 8 ESX x86 for VMware they need 600GB SAS RAID5 (ESX)
I need to create pools / Raid Groups , I will be carving Lunst later.
Now, after going through the Best Prac and few other documents and the information I have from customer - this is what I have concluded , please let me know if that is correct or what ‘re the other ways I can configure?
NO FAST /FASTVP.
I have 1 DAE which has SAS 600GB and 300 GB mixed & another which has SAS and NL SAS mixed.I know I cannot mix 300gb and 600gb disks in the same raid group or pool.That means, I need two separate pools . one 300GB and another 600GB. Raid 5
Also, HPUX needs RAID 6 2TB NL SAS. I need to create a separate pool NL SAS pool ( I should not mix SAS /NL SAS as no FAST enabler present ) .
Now with oracle as one of its applications. If I ask the customer to keep Data Luns ( application file system )in one pool , 300GB R5 pool in my case.
How do I separate the Log Luns ( means of recovery) which ‘re write intensive. If I follow the best practice I need to create Raid 1/0 Raid Group (RAID10 vs RAID5 -- logs are nearly 100% write. RAID10 has less penalty than RAID5 on writes) .
But , customer wants either Raid 5 or Raid 6 for their HPUX hosting oracle. What should be my correct approach in this scenario , please share some your experience.
First, there is nothing to prevent you from mixing 300GB & 600GB drives within a pool. Its just not recommended.
If the two different capacity drives get mixed in the same RAID group, the larger drives loose a lot of capacity when they get truncated to the capacity of the smaller.
If the two different capacity drive types are in separate RAID groups within the same pool, the larger drives will eventually get allocated a larger number of LUN slices. In general, you have more IOPS per slice with the smaller drives as capacity utilization gets higher.The response time of the 600GB SAS-based RAID group with a greater number of slices allocated is likely going to be higher than a 300GB SAS-based RAID group with fewer slices. (This assumes the SAS drives have the same speed.)
If the two drive types are different speeds and capacities, things get complicated. However, you likely end-up with peer RAID groups in the pool tier having different response times, where you what them to be as much the same as possible.
None of the circumstances above is ideal.
Ideally, you should have enough drives to put everything into pools that create separate performance and availability domains. EMC would like to sell you the drives to do that. However, sometimes the resources are 'just not there'. It maybe that for high availability and because there are not enough drives available to create an additional pool with the needed performance, the logs have to placed on a traditional RAID group sized to meet the IOPS of the logging workload. Or, the logs must be placed in the pool, at the sacrifice of some availability, because you don't have the capacity to put them elsewhere.
Document your conclusions and assumptions, and make sure that your customer understands the trade-offs that need to be made. (Maybe you'll get a larger budget next time.)
Thanks for the explanation . I am getting better now.
So can I have 3 pools created ? I understand having three is more fragmented .
(HPUX)-
1 Pool -300GB RAID5 &( for Oracle and custom applications)
1 Pool - 2TB RAID6 ( for their archiving data )
VMware -
1 Pool 600GB = RAID5 (ESX)
problem is they have - 92 +4 ( vault) SAS 300GB - so if I create 3 HS i am left with 89 disks .
so, 17*(4+1) + 1*( 3+1)
similarly , 30 600GB SAS -so if i have 1 HS i am left with 29 Disks.
so, 5*(4+1) + 1*(3+1)
Both of which deviates from Best pract.
Is there a way i can get the math right so that i don't deviate from Pool creation BP?
also,If i go for a traditional RG for Logging how many disk i can ideally use in my scenario to fit the calcuation( 300GB SAS) ? I 'm trying all alternatives.
I understand this mainly depends on the IOPS .But, the customer doesn't have a detailed information. So i would like to know in general ( with out looking into acute performance ) my aprrocah with a 3 pool is correct or not ?
My concern is with 89 and 29 SAS( 300 / 600GB) R5 drives each in a pool respectively.Please suggest . Thank for your time!
jps00
2 Intern
•
392 Posts
324
0
Posted November 22nd, 2011 05:00
Hi Deepat.
First, there is nothing to prevent you from mixing 300GB & 600GB drives within a pool. Its just not recommended.
If the two different capacity drives get mixed in the same RAID group, the larger drives loose a lot of capacity when they get truncated to the capacity of the smaller.
If the two different capacity drive types are in separate RAID groups within the same pool, the larger drives will eventually get allocated a larger number of LUN slices. In general, you have more IOPS per slice with the smaller drives as capacity utilization gets higher.The response time of the 600GB SAS-based RAID group with a greater number of slices allocated is likely going to be higher than a 300GB SAS-based RAID group with fewer slices. (This assumes the SAS drives have the same speed.)
If the two drive types are different speeds and capacities, things get complicated. However, you likely end-up with peer RAID groups in the pool tier having different response times, where you what them to be as much the same as possible.
None of the circumstances above is ideal.
Ideally, you should have enough drives to put everything into pools that create separate performance and availability domains. EMC would like to sell you the drives to do that.
However, sometimes the resources are 'just not there'. It maybe that for high availability and because there are not enough drives available to create an additional pool with the needed performance, the logs have to placed on a traditional RAID group sized to meet the IOPS of the logging workload. Or, the logs must be placed in the pool, at the sacrifice of some availability, because you don't have the capacity to put them elsewhere.
Document your conclusions and assumptions, and make sure that your customer understands the trade-offs that need to be made. (Maybe you'll get a larger budget next time.)
HTH