September 8, 2011 at 7:16 pm
I am currently trying to figure out what would be the best configuration for a new SAN for SQL Server 2008 Cluster. Right now I have a SQL cluster that is using a LUN on my companies existing SAN but the performance just isn't cutting it anymore and they have allowed me to purchase a new SAN, however I need to come up with the Hard Drive configuration first (we don't have a real SAN Administrator). Based on some articles I have read (I have been at this for 2 days) I have come up with the following configuration.
14 Drive SAN:
Virtual Disk 1 (6 Disks): RAID 5 - Data files
Virtual Disk 2 (2 Disks): RAID 1 - Log files
Virtual Disk 3 (2 Disks): RAID 1 - Temp DB
Virtual Disk 4 (4 Disks): RAID 5 - Backups
Can anyone offer any advice on this configuration? If it is good or bad, or if anything should be changed. Also, I still haven't figured out where I can pull the Quorum and MSDTC shared drives from. I was thinking I would carve up one of the virtual disks noted above into 3 LUN's, but I am not sure which one would be best for that.
Also, since we are going to present the new SAN to the Cluster through iSCSI I was thinking about puting 2 NIC's in each of my server nodes and have one NIC for transactions between SQL and my application servers and the second NIC for connectivity between SQL and the SAN. Is there any advantage to configuring my new SQL cluster in this manner or will I gain nothing from it? We have multiple VLAN's in my office and usually SAN's are put on the management network and the servers on a server network. The VLAN's seem like they are used more for device organization since the devices can talk to each other across the VLAN, however I am not a network guy so I don't know all the intricacies of what my networking team is doing.
Thank you for any input anyone can provide.
September 9, 2011 at 1:55 am
Actually, there is no way to tell. It all depends on your requirements. I will point out a few things though: Even with two 1Gbps ports for iSCSI, you have a limited amount of bandwidth, which I guess is shared across all your LUNs. (actually approximately half the bandwidth of USB 2.0).
Also, keep in mind that a 4 disk RAID 5 set has approximately the same write performance as a single drive on random writes, sequential may be better depending on your controller. A 6 disk RAID 5 set will only have a approximately 1.5 times the write performance of a single drive (still random). Is this acceptable? If not, I would look at RAID 10.
September 9, 2011 at 1:59 am
bsegermeister (9/8/2011)
Also, since we are going to present the new SAN to the Cluster through iSCSI I was thinking about puting 2 NIC's in each of my server nodes and have one NIC for transactions between SQL and my application servers and the second NIC for connectivity between SQL and the SAN.
you'll need 2 NICs as the absolute minimum to achieve this, in fact you should probably look at having 4 NICs to provide failover for the public connection and the iSCSI network, these should\must be separated.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 9, 2011 at 2:01 am
bsegermeister (9/8/2011)
Also, since we are going to present the new SAN to the Cluster through iSCSI I was thinking about puting 2 NIC's in each of my server nodes and have one NIC for transactions between SQL and my application servers and the second NIC for connectivity between SQL and the SAN.
you'll need 2 NICs as the absolute minimum to achieve this, in fact you should probably look at having 4 NICs to provide failover for the public connection and the iSCSI network, these should\must be separated.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 12, 2011 at 10:38 am
OK, I have reconsidered my configuration to be the following:
16 Drive SAN:
Virtual Disk 1 (8 Disks): RAID 10 - Data files
Virtual Disk 2 (2 Disks): RAID 1 - Log files
Virtual Disk 3 (2 Disks): RAID 1 - Temp DB
Virtual Disk 4 (4 Disks): RAID 5 - Backups
My databases are a mix of high read and high write databases, so I am hoping RAID 10 will be sufficient for both.
Also, as far as the Quorum and MSDTC drives is there any negative affects to carving them out of Virtual Disk 4, which is part of the drive I am going to use for backups.
September 12, 2011 at 10:50 am
bsegermeister (9/12/2011)
Also, as far as the Quorum and MSDTC drives is there any negative affects to carving them out of Virtual Disk 4, which is part of the drive I am going to use for backups.
you could use a majority node set quorum with a file share witness to get round the requirement for a quorum drive. More on this at this link
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 12, 2011 at 11:02 am
Perry Whittle (9/12/2011)
bsegermeister (9/12/2011)
Also, as far as the Quorum and MSDTC drives is there any negative affects to carving them out of Virtual Disk 4, which is part of the drive I am going to use for backups.you could use a majority node set quorum with a file share witness to get round the requirement for a quorum drive. More on this at this link
I did not know about that. Can something similar be used for the MSDTC, or does that have to be a shared resource?
September 12, 2011 at 11:17 am
File share witness for Maj Node Sets is a little known trick in Windows 2003 but works well and is a standard quorum option in Windows 2008.
The DTC data must use a shared disk.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply