Contoso Labs Series - Table of Contents

We've got our goals, users, and services roughly defined already, so we need to start figuring out what resources we need to actually build this cloud.

What We've Got

As an existing asset base, we have 288 HP Gen6 servers. While certainly not the newest hardware available, they're more than adequate as VM hosts for our purposes. Most critically from our standpoint: They're paid for! This project does not have a bottomless budget, and they will provide not only compute for the cloud, but the VM hosts that the management VMs will run on, as well as provide the network edge services for our NVGRE VM networks. These systems literally represent millions of dollars of paid-for assets. No way we're looking that gift horse in the mouth just because there's newer stuff out there to buy.

What's Missing

First: Network equipment. Normally, our own IT group provides things like network services to deployments in our Tier-2 datacenter, but since we're acting as our own external entity that just happens to be co-located inside the DC, we need to bring our own network and network equipment. That means everything from the network edge, to the top-of-rack switches, as well as any ancillary functions we might need like load balancers and firewalls.

Second: Storage. We need a place to keep all of these virtual machines we're going to end up running. While each of our nodes has ten 600GB 10k rpm SAS disks inside them, they obviously don't provide for any form of continuous- or even high-availability. We need some storage that can be used by the various cloud services we'll be running.

Miscellaneous: Cabling, adapters, peripheral updates, etc. Putting this here as a reminder that these 'incidental' costs rack up quickly when looking at large scale deployments. Cabling for 10Gbit Ethernet, high-speed external SAS, and other connectivity in an advanced fabric tends to add up. Even with the discounted purchasing power of Microsoft behind us, this is adding up to tens of thousands of dollars.

In subsequent posts, we'll go into all of the above topics in more (often excruciating) detail. Next however, we'll tackle the high-level fabric architecture we've decided on, and how we made those choices.