Even in a recession you can find businesses that are growing and administrators who must expand their infrastructure...
to match their company's needs. What happens when you need to expand in a facility but you've run out of space? These days, even if your organization is growing, there's a greater push to grow prudently and efficiently while saving costs, if possible. Traditionally you might expand your infrastructure by growing into a colocation space or building out a new data center, but if you want to cut costs you might want to look into some ways to make the most out of the space you already have.
In this article, I'm going to discuss a few techniques to stretch your current data center a bit further. Most of these suggestions follow common sense and are based on the assumption that you've been in your current location long enough for some of your equipment to show its age. Even if you have plenty of space right now, these ideas can still help you make efficient use of it.
Consolidate outdated servers on new hardware
The first and most obvious way to get more capacity out of your current space is to replace older, slower machines with newer, faster ones. These days, with dense, multi-core processors and RAM prices always seeming to fall, you might find a 1U or 2U server that can match or exceed the power of the 2U or 4U server you bought a few years ago. With the advent of small, serial-attached SCSI drives on rackmount servers, you probably won't have to sacrifice storage capacity in your new, smaller server.
Depending on the age and capacity of your older gear, you might find that you can combine a few older servers onto a single modern machine. Those two old CPU systems might combine well into a modern quad-core machine with plenty of RAM. In addition to saving space, you might even find that the services run faster on the new gear.
It's amazing how fast technology prices fall. I remember when we originally priced out a particular set of servers, we discovered that it was actually cheaper to buy two servers with half of the maximum RAM capacity than one server at full RAM capacity. Of course, a few years later those DIMMs are no longer state of the art, so I can now get twice the machine for the same price. Or, if I want to save a few bucks, I can replace the old RAM with larger -- and now cheaper -- DIMMs.
Deploy sever virtualization
The ultimate way to consolidate older servers is to put them in a virtual container. There are many articles that tout the benefit of virtualization in the data center, so I won't rehash the arguments here, but suffice to say that in many cases, servers don't use 100% of their hardware resources. With server virtualization, you can combine many servers onto one or more virtual machines and get the most efficient use of the hardware, not to mention power savings and administrative advantages. Depending on the application, I've found VM-to-server ratios ranging from 4-to-1 to 12-to-1 -- that's a lot of freed up space.
Use built-in remote management
Even though we've had built-in remote management on servers for years, I still see administrators rack brand-new gear that's managed by old (and large) serial consoles and remote power units. Not only are most vendors' remote management features far more capable and simpler to set up and use, they take up no extra space or power in the rack. Expand services in the cloud
As with virtualization, cloud computing gets lots of buzz these days. Depending on your infrastructure, though, it might make sense to do at least some of your expansion through cloud computing or cloud storage. This might even work well as a stopgap solution until you have time to free up enough space in your current footprint.
Invest in blade infrastructure
Blade servers are purpose-built to help you consolidate a large amount of hardware in a small space. For one, you can typically fit more blade servers into a space than 1U machines. For instance, one vendor squeezes 16 blade servers that are as capable as some 1U or 2U counterparts into only 10 rack units of space. On top of that modern blade chassis, consider incorporating network and SAN gear into the same space, so on top of those 16 servers you might also fit a few network and SAN switches in the same 10U.
If you have already invested in blades, you still might be able to squeeze more space out of them if you aren't using all of the network and SAN integration. Moving to integrated network switches instead of the cabinet switches common in many racks might save you two, four or even more rack units per rack, not to mention a ton of cabling headaches.
Whether or not you need to reclaim space in your server room, consolidation and efficient use of rack space is a topic of many discussions between data center administrators and vendors. Not only will most of these space-saving solutions cut down on power and heating costs, but many of them will also greatly simplify your day-to-day server administration and future expansion.
ABOUT THE AUTHOR: Kyle Rankin is a systems administrator in the San Francisco Bay Area and the author of a number of books including Knoppix Hacks and Ubuntu Hacks for O'Reilly Media.
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at firstname.lastname@example.org.