Hardrive recommendation for SQL Server 2008/web server

[brandonbull]

I'm trying to gather supporting facts for our reporting group's hardware recommendation. They requested either 10k or 15k drives to support an environment that would have SQL Server 2008 / SAS / Report services hosting.

There would be 3 VMs installed on the hardware, 1 for "dev" and 1 for "prod", and 1 for SAS. The "dev" environment would support development and ad-hoc sql requests and "prod" would support production reports, ETLs, and a small ASP.net app . The original request was for 2 seperate boxes but IT is pushing back with 1 box and 7.2k drives in RAID 1+0.

I need to find, if any, supporting documents for the 10k/15k server because it would be reading/writing from the same set of drives.

 

[somethingsketchy]

If you are going to have thousands to tens of thousands of queries, sql requests, etc., then 10k HDDs are worth it, but if you are only developing for a small scale application 7200rpm HDDs are okay. Depending on the volume of users and the volume of requests each user makes, your report will have to cover that.

Before you delve too much time and energy into this report, I would see what kinds of requirements the reporting group is seeking. Their answer will determine if you will recommend 10-15K rpm HDDs or 7200rpm HDDs.

 

[brandonbull]

The web app/reporting services will need to support 50+ reports for around 5-10 internal users and several external clients.

The hardware will need to also support 3 data analysts for the marketing/analytics team.

 

[brandonbull]

Any suggestions?

 

[supertle55]

Since you are going from 2 separate servers and being crammed into 1 Server to run mutiple VMs, I recommend you stick with the 15k HD. Check with capacity needed as the 15k are much smaller to support VM OS's and your databases.

The increased in hardware cost is not much in comparison to service/support cost required when you suddenly find out your server is a bottleneck. Also remember to build for the future and about possible increased usage of the server. Tell them don't be penny wise and pound foolish. Saving $1k to $2k in hard cost today may mean $20k cost (or much more with loss productivities included but would be very hard to measure) in soft cost such as doing another assessment later on to upgrading servers/software so soon.

You should not get your "IT" dept involved unless they themselves are directly involved in using this server. Spec it out and send the quote to the project manager / or client.

 

[brandonbull]

Originally posted by: supertle55
Since you are going from 2 separate servers and being crammed into 1 Server to run mutiple VMs, I recommend you stick with the 15k HD. Check with capacity needed as the 15k are much smaller to support VM OS's and your databases.

The increased in hardware cost is not much in comparison to service/support cost required when you suddenly find out your server is a bottleneck. Also remember to build for the future and about possible increased usage of the server. Tell them don't be penny wise and pound foolish. Saving $1k to $2k in hard cost today may mean $20k cost (or much more with loss productivities included but would be very hard to measure) in soft cost such as doing another assessment later on to upgrading servers/software so soon.

You should not get your "IT" dept involved unless they themselves are directly involved in using this server. Spec it out and send the quote to the project manager / or client.

We did send some specs over but IT is pushing back about using 15k drives.

I spoke with the IT director and he said he wants to use 7.2k drives to start and if we "need" more performance he could repurpose the 7.2k drives and migrate us to the 15k ones.

I just think the 7.2k drives will take a beating trying to support 4 data analysts running SQL Server queries, Report dev, ETL jobs on 1 VM, and a VM running SAS. I think the "Prod" environment will be moved to another VM in the second datacenter.

 

[VinylxScratches]

I/O on VMs is gonna be rough on the hard drives. I would say 15k now so you have less headaches later. Is there that big of a price difference?

 

[brandonbull]

Originally posted by: VinylxScratches
I/O on VMs is gonna be rough on the hard drives. I would say 15k now so you have less headaches later. Is there that big of a price difference?

I think it's around $400/drive difference between 7.2k and 15k

 

[supertle55]

I'll say it again. Hardware is cheap compare to labor. Ask him how many hours will it take to "re-purpose" a drive. Include your time and the developers time to rebuild the system on a new drive / system. Just thinking about it and going back and forth to save a grand already cost your company a grand in "reviewing the specs". Time is money. If it's not, then your IT dept has too much time on it's hand and maybe they need to cut back on their labor cost.

 

[Blain]

You keep saying that "IT is pushing back".

Who makes the final hardware purchasing decisions there?

 

[brandonbull]

Originally posted by: Blain
You keep saying that "IT is pushing back".

Who makes the final hardware purchasing decisions there?

We are trying to get a project manager involved so we can avoid IT making the final decisions. The purpose of the hardware/software is to create a reporting / analytics server.

 

[Blain]

Originally posted by: brandonbull

Originally posted by: Blain
You keep saying that "IT is pushing back".

Who makes the final hardware purchasing decisions there?

We are trying to get a project manager involved so we can avoid IT making the final decisions.

Why shouldn't IT have the final say in the purchase, since they're responsible for the system?

 

[CoinOperatedBoy]

15K would make more sense for all that I/O. But if IT is insisting on 7200rpm drives and they say you can replace them if performance is subpar, why not just do that? That seems reasonable.

 

[imagoon]

Originally posted by: supertle55
Ask him how many hours will it take to "re-purpose" a drive. Include your time and the developers time to rebuild the system on a new drive / system. Just thinking about it and going back and forth to save a grand already cost your company a grand in "reviewing the specs". Time is money. If it's not, then your IT dept has too much time on it's hand and maybe they need to cut back on their labor cost.

Why would you bother rebuilding it. It only takes like 30-45 seconds a gig to move it around the VM server farm to a machine with more guts.

-edit-Actually with vmotion you could do this with any downtime at all-edit-

What the IT guy is suggestsing is very valid. Getting all worked up about hardware is pretty poor when the IT guy can throw together a VMware box and add it to the farm in 10minutes. The IT guy also may have a much better understanding of the loads they are looking at.

Also the size of 15k drives is lower and the cost higher than 10ks and by far 7200RPM drives. 50 users may not represent much load at all. Without knowing load patterns and the type of queries that are being done it is hard to estimate this.

 

[Mide]

You'll be spending a bundle more for the faster drives. What RAID will you be running?

 

[brandonbull]

Originally posted by: Mide
You'll be spending a bundle more for the faster drives. What RAID will you be running?

RAID 10

 

[Blain]

What RAID card does the IT push?

 

[brandonbull]

They use Dell 2950 IIIs for the servers. Currently most our servers use 7.2k drives and the only the main production database and web server use 15k drives.

We are trying to set up an environment to support the reporting and analytics team. Currently manage around 150GB data for multiple clients in Postgres and Oracle environments.