Every data center will periodically enter a refresh cycle for its servers and other critical hardware. This allows organizations to deploy updated equipment intended to improve reliability, enable new capabilities and ultimately save the business money.
Traditional refresh cycles took place about every five years, but they have been accelerating over the last decade. Is it time for you to refresh servers now? The answer isn't always simple, but there are important factors that help justify an early system update.
Look for servers that enable new features and functionality. Consider a server refresh when a new make or model may enable capabilities that were difficult -- or even impossible -- with your existing servers. Remember that just a few years ago, server virtualization was cumbersome on CPUs prior to the release of Intel VT or AMD V extensions, so updating to servers with CPUs that supported virtualization extensions allowed dramatic improvements in virtual machine (VM) performance.
These capabilities are not just driven by advances in the processor. Memory is a crucial feature of servers used in virtual environments because virtual machines are essentially disk images that reside in server memory. More memory is vital for higher levels of consolidation, and the reliability of that memory will impact the overall reliability of all the VMs on that server. So, memory features may include advancements to error-correcting code memory that can extend multibit errors on a single memory device across an entire bank of memory devices, the addition of spare memory (memory sparing) that allocates extra memory devices as spares in the event that a memory device becomes unrecoverable, and the inclusion of fault-tolerant memory mirroring which duplicates memory contents across devices (similar to RAID 1 for memory).
Future capabilities may include support for new chipsets that can handle additional memory types, faster I/O, higher bus speeds and so on.
Evaluate the economics of system lease or rentals. Many companies choose to own their own servers, but there are also many different lease or rental options that can allow organizations to acquire newer and more powerful hardware while converting the associated capital expenses to operating expenses. Similarly, vendors may offer alternative payment models (such as low-interest financing) to qualifying businesses.
Companies may be hesitant to accelerate hardware refresh cycles because they are unwilling (or unable) to provide the more frequent infusion of capital needed for acquisition. In addition, the company doesn't want to bother with recycling or disposing of the displaced hardware on a more frequent basis. Nontraditional lease or rental arrangements can relieve the capital burden and solve the end-of-life disposal problems at the same time.
Look for servers that enable additional computing capabilities. Server virtualization has driven a revolution in processor design and computing capabilities, and consolidation projects and desktop virtualization initiatives may benefit from faster server refresh cycles when the investment allows greater levels of consolidation. For example, servers with Intel E7-2803 Xeon processors allow for six cores and 12 threads per processor, but updating that server with Intel E7-8870 processors provides 10 cores and 20 threads per processor. These capabilities are further multiplied when acquiring new servers with two or four processor sockets.
Think about improvements in system distribution and cooling. Congested data centers or facilities grappling with thermal problems may accelerate a server technology refresh because new servers often provide more capability in a smaller form factor. For example, the computing power contained in your current 4U rack server may now be provided in a 1U rack server.
The move to fewer, smaller form factors may allow data center operators to redistribute the server hardware across available rack space. This can improve cooling air flow patterns and alleviate equipment "hotspots." This can lead to lower cooling costs, especially if you're deploying point-cooling devices and leveraging higher data center operating temperature recommendations in line with prevailing American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) standards. In addition, eliminating hotspots can enhance system reliability, leading to lower downtime and service disruption.
Look for return on investment in power savings. Power costs now make up a major portion of a data center's operating expenses, so any effort to mitigate power costs can have a beneficial impact on the company's financial position. Consider the effects of server consolidation. By replacing older, less capable servers with newer and far more able systems, the business can perform the same amount of computing with only a fraction of the total number of systems. Fewer physical boxes demand less total power -- an operational savings that pays the company back every month. The move to capable next-generation servers can potentially save enough power to actually pay for the technology refresh.
One of the downsides to any technology refresh is removing displaced equipment. Fortunately, virtualization abstracts the workloads from the underlying hardware, thus largely overcoming traditional hardware dependence. This provides additional options for older equipment. For example, older systems can be reallocated to test and development, moved to workgroup or departmental computing tasks, or reallocated to branch office and remote data center locations where the servers can continue to operate in a secondary role. These options can keep equipment in the enterprise longer, even though it may be out of production.