Planning Your Data Center Strategy

By Adam Wallace  |  Posted 2011-10-03

Enterprises looking to revamp their strategies around data center optimization and utilization face the initial decision of whether to build from scratch or to retool their existing systems.

Creating a new data center from scratch allows an organization to plan for the greatest amount of customization. However, the significant investment in time, construction, hardware and support is prohibitive for many enterprises. Reconfiguring an existing data center is efficient and cost-effective, but it cannot always fully address an enterprise’s requirements.

See also: Extreme Makeover for Data Centers

After an enterprise decides on an initial “build or retool” approach for its data center, management must determine whether to keep the data center in-house or to outsource it to a hosting company. Businesses can choose to keep all processes in-house, thereby exercising total control and decision making on construction, implementation and maintenance.

The outsourcing model (commonly called cloud computing) reduces data center costs, much of which are attributed to the resources needed to run the data center on an ongoing basis, particularly when it comes to high-end technology expertise. This model also allows businesses to pay only for the CPU cycles needed on a per-user/per-month basis.

Before partnering with a hosted data center, enterprises should ask the cloud provider the following questions:
• Can you demonstrate a high level of security?
• What level of throughput can you provide?
• How quickly can you provision new bandwidth?
• Can you guarantee connectivity?
• What is your failover plan?

Technology Advances and Strategies

Enterprises that decide to implement a new data center strategy in-house should consider the following technology advances:

Virtualization: Virtualization software lowers power and cooling costs by reducing the number of physical servers by a factor of 5X to 10X. And with CPUs becoming more powerful, even greater workloads can be added onto each server CPU. Virtualization also significantly reduces capital expenditures because fewer servers are needed.

High-Speed Backplanes: Data center bandwidth and throughput performance have improved with higher speed backplanes. Ten gigabit Ethernet is now the standard, and, with blade chassis design, servers can communicate with each other in the same chassis so that data does not have to travel as far.

Virtual Switches: Virtualized switches enable more flexibility because businesses can deploy a server and give access to any network, rather than worrying about whether a particular network is wired to a particular server. Virtualized switches do require additional set-up time, but they simplify switch and server management and allow for quick deployment of subnetworks.

Active Disaster Recovery Data Centers: In the traditional model, a business deploys a disaster recovery data center that stores replicated data and is used only if the primary data center goes down. Today, more businesses use two data centers in tandem to share the computing services load. Both provide data and applications to users, while replicating data back-and-forth to each other. If either data center goes down, the other one can take over with a complete copy of the data to keep the business running while the downed data center is brought back online.

Low-Power Memory and CPUs: Today’s data center hardware uses significantly lower voltage, which means that servers, switches and other hardware components consume significantly less power.

Pod Designs: Previously, businesses had to cool their entire data center. Now, many businesses employ pod designs to reduce power and cooling requirements. The pods contain all the heat on the back side where exhaust fans push it outdoors. The pods then absorb ambient-temperature air through the front to keep hardware at appropriate temperatures.

Blade Technology: Blade technology features higher CPU density within a smaller physical footprint, requiring less rack space and allowing businesses to deploy more CPUs in a smaller area. The combination of blade technology and virtualization allows businesses to reduce server density significantly.

Converged Networking: With converged networking technology, a single piece of hardware can fulfill multiple functions, including high-speed networking, data transfer and storage. This also allows businesses to virtualize network-interface component addresses and server components so networks can be added on the fly.

Adam Wallace is a senior infrastructure consultant for OpenSky, in Tolland, Conn. The firm's services include infrastructure; IT risk management and security; governance, risk and compliance; and technical business consulting.