Data centers have been around - well, essentially forever, or at least since the dawn of computing. But, just because they've been around forever doesn't mean they've endured unchanged. In fact, we're smack in the middle of a radical shift in data-center architecture, design and operations.

From January to March of 2006, Nemertes Research conducted research on the transformation of data centers. We completed detailed discussions with 82 IT executives from 65 companies, spanning various industries and company sizes. The average revenue of the participants is $5 billion, the average IT budget $145.5 million. Participant companies had, on average, 24,000 employees in 385 locations. To serve these employees, companies had an average of eight data centers.

Some historical perspective: In the early 1990s and again in the early 2000s, data-center technology underwent some significant transformations. In the '90s, the Internet created new data centers for e-commerce and Web-based applications. The data center of the early Internet era was dominated by sprawling commodity servers, either "white box" systems or 1U rack mountable servers, also known as "pizza boxes." As servers proliferated, so did capacity problems for cooling and power systems. A second important transition in the late '90s, and continuing today, has been the development of more and more dense computing architectures.

As a result, many companies found that their existing power and cooling systems could not support the next generation of server technology - so unsurprisingly, our research participants cite facilities upgrades as a top-funded initiative for 2006.

Our research uncovered four major themes in data centers:

Consolidation - Whether due to regulatory compliance or the increasing cost and complexity of IT, companies from all industries and all sizes are consolidating and centralizing IT resources, including data centers. Specifically, IT managers are consolidating dozens of data centers into a few central locations - then enhancing the capacity of those few across all areas: networking, redundancy, computing, storage and management. This trend has some significant corollary impacts. For one thing, data and applications are increasingly delivered to remote users over the WAN (as our research shows, as few as 11% of workers work at headquarters), which has major impacts on WAN architecture and management. For another thing, centralized data is more easily backed-up, made redundant and controlled for compliance, which is sparking major initiatives in all these areas. Finally, centralized applications, delivered over the Web, client-server protocols or thin clients, are becoming the norm, as they are easier to manage, update and patch against security exposures.

Growth - Demand for data-center resources such as servers and storage has generated tremendous growth. The primary driver is in the growth of applications: the average enterprise has more than 1800 applications on its network and increasing at more than 10% annually. Additionally, companies are retaining more data than ever before and increasing availability by deploying redundant servers. As a result, servers are growing 11% per year on average, while storage is growing at a median rate of 22%. Rapid growth is causing strain in data centers, with insufficient capacity in environmental controls (cooling, humidity controls) power supplies and floor space. Companies struggle to balance between large footprint (sprawl of low-density servers) or high density (super-hot and power-hungry, dense racks).

Availability - Business demand for higher availability is increasing across all industry verticals. Half of all participants list availability as one of the top challenges. Correspondingly, almost half of the research participants are setting up secondary data centers for continuity and disaster recovery, either by repurposing an existing data center, or by building new facilities in which they consolidate existing data centers. After Hurricane Katrina, disaster recovery consultants and managers have re-evaluated traditional assumptions about the potential geographic footprint of a disaster. As a result, secondary and tertiary data centers are often built more than 200 miles from the primary data center.

Operational Efficiency - Companies are focusing on data-center management and operations, attempting to cut costs, improve efficiency, automate repetitive tasks and better align IT spending with business needs and demand for services. More than three-quarters of respondents take a holistic view of the data center as part of a broader shift in IT culture to "IT as a service." This translates to data centers as business-service delivery centers. The emphasis on business services extends to broad adoption (more than 50%) of governance and service delivery standards such as IT Infrastructure Library (ITIL®) and Control Objectives for Information and related Technology (COBIT®). Only a few companies are implementing these standards formally and in their totality, but even partial implementations can bring benefits in change management practices and, as a result, in higher availability and lower operational costs.

Conclusions

Here are some recommendations that follow from the trends Nemertes sees in the data center.

• Increased bandwidth and protocol-optimization appliances have made it possible to consolidate many of the servers located in smaller offices into centrally located data centers. Consolidation can reduce the need for IT staff at many locations, reduce management and operations costs, and facilitate compliance with regulations.
• Many servers operate at very low levels of CPU utilization. They still carry the full costs of maintenance and software licenses. Using virtualization software, IT managers can consolidate many underutilized servers onto one physical server. The savings from postponing new server purchases, re-deployment of the freed servers can be significant.
• Most companies are also consolidating storage resources. By creating a single, large pool of storage, companies can reduce management costs and increase the utilization and availability of storage resources.
• Servers are becoming faster, more power hungry and hotter. Companies are upgrading cooling and power facilities in data centers to meet the increased demands. Data-center design is complex - IT managers should consider consulting a specialist, consultant or vendor to help plan facilities upgrades.
• Nemertes recommends that IT executives create a position of data-center architect, a senior role with planning and strategic responsibility across all the various silos and disciplines in the data center: servers, networks, storage, management, monitoring, security and facilities.
• Increased demand for availability is pushing companies to spend more resources on data center redundancy. Network resiliency is now even more important as the WAN is used to deliver applications to remote workers. Companies should re-evaluate their availability and BCP/DR strategies in the context of consolidated data centers and dispersed workforces.

Andreas M. Antonopoulos is SVP and Founding Partner at Nemertes Research Inc. (www.nemertes.com). He may be reached at andreas@nemertes.com.