Influencing the infrastructure of the data center

Product cycles have been condensed from years into weeks, and supply chains are interwoven into a complex fabric. Paper is giving way to pixels and e-business is rapidly evolving from an elusive dream into reality.

In this post-industrial environment, an enterprise that has a well-defined data management strategy is indisputably at a tremendous advantage. Such an organization can harness the power of information technology to achieve remarkable results. Conversely, companies that lack the right server and data management strategy risk missing out on the promise and opportunities of the digital age.

Achieving success is no simple task. As organizations have adopted a heterogeneous mix of hardware and software systems—with servers scattered across an enterprise—the solution has often created a new problem. Today’s computing environments have grown exponentially, leaving companies with larger and more rigid data centers. Not only does it become cost-prohibitive to administer these data centers, it is difficult to achieve optimal value and return on investment as well as return on assets. Hard-to-find data, old and inaccurate data, aging hardware and software and sluggish performance can all conspire to drag down service levels and profits. In the end, an organization can find its customers, business partners and employees dissatisfied.

There are two trends in the computing industry that promise to have significant influence on the underlying infrastructure of the data center – virtualization and grid computing. Neither of these are new technologies, but the application of these technologies with x86 processor-based systems is having a rather profound impact in the way our data centers are being designed and managed. With computing demands increasing every year, and as much as 80 cents of every IT dollar spent on software and operations maintenance, the need to streamline has become paramount.

Virtualization is a technology that is a standard feature in the mainframe and UNIX world and is now an emerging technology for x86-based computers. Enterprises have been using virtualization technology on mainframes and RISC-based systems to enable better utilization of hardware resources. As x86 servers have become a mainstay in the enterprise, more companies are exploring virtualization with these servers to enable more productive, flexible and scalable data centers while reducing costs and boosting data availability. 64-bit and 32-bit computing technologies from AMD are providing the foundation for today’s— and tomorrow’s—enterprise virtualization solutions.

The goal of virtualization to break the ‘one server, one application’ model has been the traditional approach to date. As x86-based servers have become a dominant force in the data center, this approach has resulted in a growing number of highly capable servers that are drastically underutilized. We have customers who are telling us their utilization rates for their x86-based servers are under 20 percent. These servers take up space, put strains on power and cooling systems and demand more staff to monitor and manage them.

Virtualization technology allows you to either run multiple virtual machines, with their own separate operating system and application, on one computer or to combine several servers together to make one large virtual machine. Either way, the goal of virtualization is to spare the user from having to understand and manage complicated details of server resources while improving operational efficiency, increasing resource sharing and utilization, and maintaining the capacity to expand later.

Some of the benefits that virtualization can provide to an organization include:

• Consolidation of workloads of underutilized servers to help save hardware, power, cooling, and space costs
• Run legacy applications on newer and supported hardware and operating systems
• Run multiple operation systems simultaneously on one system, which helps reduce the amount of hardware needed to develop and test new applications and test and deploy new configurations
• Provide a more secure and fault tolerant environment since the functioning of one virtual machine does not necessarily affect the functioning of another
• Help make tasks such as system migration, backup, and recovery easier and more manageable.

Since virtualization is an emerging technology in the x86 world, there are issues that need to be addressed as companies begin to deploy. There is some overhead associated with virtualization, so heavy-use production applications might not be the best candidates. For example, we don’t necessarily see production-level databases migrating to virtual machines. Also, there is a lot of activity in the area of virtualization software as established players like VMware and Microsoft continue to mature their codes and newer players, like the Xen open source hypervisor mature code.

Implementing virtualization requires careful thought in order to help reduce management issues. Companies need to take a close look at their current computing resources, decide the goal they want to achieve with virtualization, and pick “best candidate” workloads for virtualization. You want to avoid exchanging physical server sprawl for virtual server sprawl. You also need to have in mind the management tools you will use with your virtual environment.

We initially see virtualization being adopted by many companies to help consolidate workloads and to address the power, cooling, and space issues facing most enterprise data centers. In general they start this consolidation for specific types of applications or workloads, to perhaps consolidate legacy business applications. They then apply the value of virtualization to business continuity, enabling a redundant infrastructure that is based on less physical hardware, reducing the cost and management issues.

Once an enterprise sees the benefits of virtualization from these initial consolidation efforts – they begin to widen their view and consider virtual environments as a way to fully leverage their hardware and software assets. At this point they start to think of virtualization as a general strategy for building infrastructure. And this is when they start to think more in terms of grid computing.

Grid computing is an emerging IT infrastructure architecture that provides the ability to aggregate networked computers and storage into a virtual pool of computing and data resources. The idea is to move from today’s architecture, which is based on dedicated resources, to one of shared resources where organizations can better optimize and manage their valuable hardware investments.

Virtualization is key technology for grid computing. It provides the abstraction needed to break the physical tie of software to hardware providing the ability to have a single, transparent view of resources. You no longer have the one-to-one relationship of server to application as we see today in the data center. Instead, applications run in virtual machines, which can be assigned to run on any given hardware in the computing grid. By freeing the software from underlying hardware dependencies, you get a new level of flexibility.

Grid computing holds significant potential as a next step in the evolution of the dynamic data center. The concept of grid computing first emerged in the mid-1990s in high performance computing where grids were used to harness together the power of many computers together to solve large, complex problems. Government labs and scientific organizations have been using grid computing technologies for several years for applications such as weather modeling and weapons test simulations. At AMD, we use this type of grid computing to support the design and testing of our processors.

Today, grid computing has moved out of the scientific and engineering world and is being deployed in more traditional commercial computing applications. We see more companies turning to grid computing to help create an IT infrastructure that is more dynamic and efficient. An example of where grid computing is being applied in the commercial world today is the banking industry, which is using grids in risk management systems.

Interestingly, many industry analysts see the move to grid computing in the enterprise similar to the way the Internet moved from research and academic circles to the business world in the nineties.

The real business value of the grid lies in the ability to aggregate computing and data resources so that a company can harness these assets to accelerate different aspects of their business, such as data analysis for business intelligence, improved collaboration and providing employees on-demand access to vast IT resources.

Some of the specific benefits of grid computing include:

• Better utilization of computing resources since grid computing pools resources into a virtual pool that can then be accessed by a large number of users
• Ability to more flexibly increase or decrease computing resources for a particular application or workload to match the demand cycles of your business
• Simplify the way you select, acquire, and use your IT infrastructure, for example allowing your infrastructure to be based on more power and cost efficient industry-standard servers

There's enormous promise in taking a data center from a collection of discreet servers to a virtual pool of resources. Of course getting to these benefits is not necessarily straight forward. As grid technology gains momentum, IT professionals will need to understand the variety of options available in order to make decisions that are right for their infrastructure.

The cultural barriers to both virtualization and grid computing need to be overcome . Our IT world has been constructed into silos of hardware and the people that manage the hardware. These silos exist because networking, database, storage, servers, and applications have become increasing complex. One of the issues with the grid is getting rid of the silos.

We are only at the beginning of virtualization and grid computing with x86-based processors. Hardware-enabled virtualization, a function available in AMD’s current generation of x86-based processors, can help to reduce the complexity and overhead associated with virtualization. It provides the underlying technology that virtualization software can use to help create virtual machines and to enable more efficient switching between these environments. Hardware-enabled virtualization is in its infancy and as this technology matures we will see virtualization of devices and better optimizations that enable near-native performance of applications running in virtual machines.

The x86-based processor is a core component in enabling grid computing. The 64-bit and multi-core capabilities that have evolved in x86-based processors provide the type of computing density that is perfectly aligned to fuel the growth of grid computing. Hardware-enabled virtualization, an emerging feature in x86-based processors, is also helping to reduce the complexity and overhead associated with virtualization software.

A growing number of businesses are turning to virtualization and grid computing solutions to help them build highly flexible and scalable solution platforms. As business decision-makers increasingly look to squeeze out additional gains from x86 architecture, they are taking a closer and more critical look at the servers they purchase. Certainly, it is no longer feasible to buy a server for every application an enterprise plugs in. Managing IT resources efficiently is a necessity. Only then is it possible for an enterprise to ensure that it is maximizing revenues, return on assets, customer and partner loyalty, and shareholder value.

Organizations are discovering that it is possible to boost productivity and slash costs. Whether the focus is on system-wide performance, e-commerce, storage or simply maximizing the use of data, tomorrow’s solutions exist today. AMD processor technology is ushering in a new era of server performance and is demonstrating that virtualization and grid computing can provide a giant leap forward in the evolution of computing and information technology.