Need an opinion on SAN setup please
Posted: Fri Apr 06, 2012 5:40 pm
(
Sorry, this post is turning out to be a lot longer than i had anticipated) I had posted a while back about a SAN migration I'm in the process of completing but got side tracked with by other projects and an unusually long wait time for our SAN components. But, now I have everything built and am getting ready to deploy the new SAN's and would like to get your all's opinion on my current setup before hand to see if you all can spot any red flags.
We will be deploying 2 identical Supermicro SAN's running StarWind 5.8 (or latest build) for hosting our 4 CSV's + Quorum.
Hyper-V Servers: Dell PowerEdge T710, Server 2008R2 Datacenter, +/- 20VM's running on 4 CSV with HA targets for failover.
SAN Hardware:
4u Supermicro SC847 Chassis- 36 HDD bays
Supermicro X8DT6 Motherboard
2 Onboard Intel 82574L Gig NIC's
2 4 Port Intel 82576 Gig NIC's (2 cards, 8 ports total)
2 Intel Xeon E5640 2.67GHz processors
16GB (4x4GB Dimms) DDR3 ECC Registered memory
36 3.5" 600GB 15K RPM SAS HDD's - Seagate ST3600057SS
Supermicro SMC2108 RAID Controller (I believe this is just a rebranded LSI 2108) with BBU. 512MB memory/cache
RAID Settings/Arrays:
OS: 1 RAID1 array for OS- Windows Server 2008 R2 Enterprise
iSCSI: 1 RAID10 array- 32 HDD total, 2 Spans, 16HDD per span. 64KB Stripe size, Disk cache enabled, Cached IO, Write Back with BBU enabled 8.717TB usable space
2 hot spares
Some RAID benchmark numbers (Any thoughts on these, I don't have anything to compare these to, do the numbers look "decent" given the hardware/setup?) Benchmark Numbers using IOMeter: Max Disk size 32000000 Sectors, 64 Outstanding I/O's per target 16 Workers (Access specifications: Max Read/Write- 32KB, 100% Sequential, 100% Read and 100% Write and for "real life" tests 8KB, 40%Seq, 60%Random, 35% Write, 65% Read)
For networking I was wondering if I could get your opinion. We have a total of 10 Gig NIC ports per SAN, what's the best practice for allocating NIC's to iSCSI vs Sync? Should we team 4 NIC's for iSCSI (LACP w/ Jumboframes) and 4 for Sync or allocate more to iSCSI?
Thank you so much for any feedback and help you can offer!

We will be deploying 2 identical Supermicro SAN's running StarWind 5.8 (or latest build) for hosting our 4 CSV's + Quorum.
Hyper-V Servers: Dell PowerEdge T710, Server 2008R2 Datacenter, +/- 20VM's running on 4 CSV with HA targets for failover.
SAN Hardware:
4u Supermicro SC847 Chassis- 36 HDD bays
Supermicro X8DT6 Motherboard
2 Onboard Intel 82574L Gig NIC's
2 4 Port Intel 82576 Gig NIC's (2 cards, 8 ports total)
2 Intel Xeon E5640 2.67GHz processors
16GB (4x4GB Dimms) DDR3 ECC Registered memory
36 3.5" 600GB 15K RPM SAS HDD's - Seagate ST3600057SS
Supermicro SMC2108 RAID Controller (I believe this is just a rebranded LSI 2108) with BBU. 512MB memory/cache
RAID Settings/Arrays:
OS: 1 RAID1 array for OS- Windows Server 2008 R2 Enterprise
iSCSI: 1 RAID10 array- 32 HDD total, 2 Spans, 16HDD per span. 64KB Stripe size, Disk cache enabled, Cached IO, Write Back with BBU enabled 8.717TB usable space
2 hot spares
Some RAID benchmark numbers (Any thoughts on these, I don't have anything to compare these to, do the numbers look "decent" given the hardware/setup?) Benchmark Numbers using IOMeter: Max Disk size 32000000 Sectors, 64 Outstanding I/O's per target 16 Workers (Access specifications: Max Read/Write- 32KB, 100% Sequential, 100% Read and 100% Write and for "real life" tests 8KB, 40%Seq, 60%Random, 35% Write, 65% Read)
For networking I was wondering if I could get your opinion. We have a total of 10 Gig NIC ports per SAN, what's the best practice for allocating NIC's to iSCSI vs Sync? Should we team 4 NIC's for iSCSI (LACP w/ Jumboframes) and 4 for Sync or allocate more to iSCSI?
Thank you so much for any feedback and help you can offer!