Server consolidation input

[ochadd]

I work for a midsized company that has pretty limited hardware requirements on a per server basis. We have 6 servers at the main location with a total of 21 active spindles not including hot spares.

1 server = 3x36GB 15k RAID 5 sets (SQL server, PDC, file)
1 server = 3x36GB 15k RAID 5 sets (SQL server)
1 server = 3x146GB 15k RAID 5 set (SQL), 2x74GB RAID 1 (OS)
1 server = 2x80GB RAID 1 (Terminal server)
1 server = 4x160GB 7k RAID 5, (SQL, file, misc) 2x80GB RAID 1 (OS)
1 server = 2x80GB RAID 1 (Misc, BDC, print)

We have about 800GB total used storage. There is 14GB of memory in play, very little unused. There are 14 cores between them as well, most sitting idle.

I'm thinking of consolidating all the SQL servers to start with and possibly move most, if not all, file serving duties over to the same machine. None of the servers go over a queue depth of 1.5 and they are all running three disk RAID 5 sets with 15k rpm SCSI drives.

I'm thinking:
Dual quad Shanghai server
32GB of DDR2 400
Sata drives in 14 drive RAID 10 or 15 drive RAID 6 + hotspare
Areca controller w/ 1GB of cache
Quad port Intel NIC

I can get all this for $6000 and utilize existing licenses. This seems like it would be more than enough to replace what's currently out there and allow me 4x-6x the storage capacity. I'd do a single install with all databases on the same server.

Keep a couple servers for domain, network, and print facilities and be good for a couple years. Thoughts?

 

[dr.sandman]

Is power consumption your goal here because its not like your running a lot of machines total? If I read your post correctly you have machines that are already underutilized and you are planning on purchasing more hardware? --Just kidding - - I wish I had your problem at my work.
And you know just as the rest of us admins do, that it would take a multi paged overview of your entire system structure and usage to make an educated guess as to what your best option would be. If you have $6k burning a hole in your budget than you need to convince the boss that GTX280 video cards will help MS Word load faster so that you can all play some CallOfDuty at work.

 

[ochadd]

The problems I'm trying to solve:

Lack of storage per server. Two machines have SCSI controllers and 36GB drives. Down to about 20% free space on them. These two have plenty of IO but not enough space. They need another 3GB of memory between them.

Need more IO. The one server with 146GB drives has plenty of space but disk queues are too high during peak times. Due to inefficiencies in the software we use it's cheaper to throw hardware at it.

Need more IO and space. The server with 160GB disks chokes during maintenance activities and under a particular load that it faces a few times a year. This machine is maxed out on disks, no more room. I'm considering losing the hot spare to give it another active spindle.

It seems to me to be more efficient to build another server that should cover out storage needs for the forseeable future rather than stick a few more $ in each existing machine to hit a dead end in another year. I might be wrong. The other worry is about building vs buying.

 

[engiNURD]

If you have any free channels on any of those controllers, just throw together a SAN. If you don't have any free controllers, but one of the servers has the capability to support another SAS controller, get one and build a SAN. If no server can support another controller card, build your system or get a NAS.