I am looking into options for building a data center at a colo for the purpose of using it as a place to host shared server websites, as well as virtual private servers.

NAS/SAN solutions, load balancers, configs, brand recommendations, etc. all would be welcome.

The idea would be to maximize the dollar as much as possible for a small to med startup and on a budget. I don't need raw IOPS at the moment, but storage and reliability is a concern at this time with the ability to address IOPS at a later time with minimal further investment.

Budget to get started is about $7000-10,000. However it would be neat to see "ideal" topographies and configs if such a small budget wasn't a concern.

I think this could be useful info for the frugal start up as well.


With 7K-10K budget, I don't think you will have the option for much of a topography.

I would just start with a single Dell PowerEdge 7610 with two Xeon E5502's, 16 Gigs of RAM, the PERC SAS RAID card, and two 1 Terabyte hard drives in a RAID 1 array, and Windows 2008 DataCenter Edition x64. That will run you about $9900.

With that setup, you should be able to run between 10 and 20 Hyper-V instances, depending on the configurations of course.

Then you need to think about backups. Since your budget is tight, you might want to skip a SAS SAN box and just go for a tape backup like the Dell PowerVault 114T. That should run you a little over $1000.

Once you get some more cash in, then I would put it towards a hardware firewall then something like a PowerVault MD3000i SAN for backups, and finally something faster than RAID 1 for your server(s).

BTW...if you aren't actually in the hosting business, you might want to consider just getting a hosted server from ServerBeach. With a Hyper-V setup, it is super easy to hop to a new server when they release upgraded packages.


A pair of reasonably well-resourced servers running Linux with DRBD replicating individual LVs to provide redundancy, and heartbeat to manage failover.

As you scale up, you can spread the replication out more to spread the impact of a failure across the rest of your infrastructure. For example, if you've got 10 servers carrying 200 VMs -- so 20 VMs per physical machine -- and want to be able to survive the failure of two failures with minimal chance of losing a VM, then you spread the disk space for the 20 VMs out across the nine other machines. Coordinating this to make sure everyone's replicated somewhere and the distribution stays stable takes a bit of coding, but isn't much of a challenge.

As far as going for a SAN, I've been there and I don't like it. Sure, if you've got a quarter mill to drop you can get something that is N-ways redundant and comes with support contracts and the rest, and will probably stay up forever without a problem. However, at your sort of price point (or even at the $150k mark, which is where I was) a SAN just provides a nice, tasty single point of failure for Murphy to play with to take your gear down at an inappropriate moment. At least with replicated disk volumes, if the worst happens you can only take out a small proportion of your customer base at once.


On the hardware side you should be able to get big discounts for being a startup. Sun give huge discounts to startups and I'm sure HP/Dell etc will do the same.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy