V400 with 64 600GB FC 15k disks. We have 16 8Gb FC lines running to two separate fabrics. We have ~100 ESXi 5.1 hosts connected to all shared LUNs. There are about 12 hosts in each cluster and 4 LUNs per cluster.
I am still experimenting with RAID1 vs RAID5 and the size of datastores (512GB, 1TB, 2TB, etc) but I am just curious what kind of stats I should be seeing?
Mainly, I want to know:
1) throughput
2) IOPS
3) latency
I know it is not fair to judge everything based off of throughput, but I am really curious what others are seeing/what I should be seeing? I am trying IOMeter tests and doing MAX THROUGHPUT and am thinking something must be wrong...
What kind of throughput and performance should I be seeing?
Re: What kind of throughput and performance should I be seei
That is hard to predict as workload will have an impact. I can tell you we have V400 with 384 300GB 15K drives, and 128 2TB drives (recently added 128 900GB 10K drives) and our array is designed fro around 80K IOPS. First I encourage you get get and install System Reporter it is a great tool for visibility into the array.
We sustain upward of 30K IOPS all day long with average of well under 10ms latency with a fairly consistent 500MBs/sec throughput (see note below)and during backups at night run 40-60K IOPS and 1-2.5GBs/sec.
Your HP team should be able to give you hypothetical numbers on your configuration, but as usual your mileage may vary
The note I refer to is System Reporter will give you different numbers depending on what you run the report on. I run my high level report against all VVs and hosts, which tends to level out the peaks. I do then run specific reports agains front end ports that are dedicated to specific functions (VM Hosts, SQL, AIX) to get better insight into each platforms performance. I then agains run more detailed reports against volumes for specific applications so that I can demonstrate how good a given app is performing on the array. I publish the URL so that app owners can loo at the array performance and not come to me with "my app is slow it must be your storage"
We sustain upward of 30K IOPS all day long with average of well under 10ms latency with a fairly consistent 500MBs/sec throughput (see note below)and during backups at night run 40-60K IOPS and 1-2.5GBs/sec.
Your HP team should be able to give you hypothetical numbers on your configuration, but as usual your mileage may vary
The note I refer to is System Reporter will give you different numbers depending on what you run the report on. I run my high level report against all VVs and hosts, which tends to level out the peaks. I do then run specific reports agains front end ports that are dedicated to specific functions (VM Hosts, SQL, AIX) to get better insight into each platforms performance. I then agains run more detailed reports against volumes for specific applications so that I can demonstrate how good a given app is performing on the array. I publish the URL so that app owners can loo at the array performance and not come to me with "my app is slow it must be your storage"