Part one on putting together a consolidation plan for better TCO and lower management costs: I am hoping you will be able to give me an answer. I find your answers well thought out and without bias.
This case study is in a university environment. The following projects need solutions:
- Exchange consolidation from eight separate servers into one manageable solution. There are 3,500 admin users. No student storage is needed.
- Storage consolidation for W2K, Netware and Solaris onto a single manageable solution. They currently have multiple SCSI arrays and a single CPQ SAN. There are over 250 different servers all over the campus and at three remote learning centers.
- Storage management solution. There is no current management for their tape backup or archiving. Departments manage their own backups. The university has Fibre connections between all the buildings and T1s to their three remote campuses.
Please advise on a consolidation plan.
There are quite a few different methods you could use to accomplish your consolidation efforts. I'll try and impart a couple that may be of interest to you. Consolidation is one way to get a better handle on management costs and better TCO for IT solutions. Although there are always different methods that can be employed to get to a final solution, the budget you allocate to your efforts will be the limiting factor of what you can and cannot do.
Let's start with your Exchange environment:
Your question did not indicate that you have remote MTAs so I'll leave those out for simplicity. I'll stick to the basic message stores and Exchange engine.
Get some eight-way servers for the Exchange environment and create a cluster using either a Veritas or Microsoft cluster server. This will require external storage Exchange. Perhaps you can connect the Exchange servers into the existing Compaq SAN to save some money. Use two host bus adapters in each server and connect each adapter to a different SAN switch. Each SAN switch should connect to both controllers of the Compaq storage array. This environment will provide a "No Single Point of Failure" solution for Exchange and provide automatic failover and failback. Make sure your SAN storage array can do snapshots. Using the SAN storage array to take snapshots of the message stores will provide instant restore capability should you experience data corruption or invasion by a virus.
You did not mention which version of Exchange you are using, but I suggest Exchange 2000 to take advantage of multiple storage groups and message store partitions.
Since you want to make sure there are no bottlencks, use a GiG-E network for client access to these servers if you can. If not, use multiple 100base-T connections. If you are using Compaq servers get dual network cards and "TEAM" them to provide network failover and higher bandwidth. You will also require a private network for the cluster heartbeat, but a simple crossover cable can be used for the connection between the cluster server nodes.
Place your cluster quorum resource on a separate physical mirror set in the storage array as that is the only single failure point in a cluster. NEVER place any other data on the quorum disk. You should also use a mirrorset for the transaction logs and keep the logs on separate disks away from the actual message stores. The stores can be placed on a RAID5 set. All data must be located in the external SAN array so it can be accessed by both servers.
This was first published in June 2003