BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Data center hardware lasts for many years, but there's a point when new servers are the only way to restore your competitive edge or ramp up a new enabling technology.
New features make sense to the IT department. If new servers are on your 2015 IT budget, they'll likely use the Intel Xeon E5-2600/1600 v3 processor. The new Xeon E5 chips support software-defined infrastructure, optimizing server behaviors based on telemetry gleaned from across the actual data center. Xeon E5 chips also offer up to 18 cores per package, which can handle 265-bit vector integer instructions. The result is higher performance; more virtual machines (VMs) per server; DDR4 memory support; and faster, better data security.
The trick to planning a server refresh is knowing where your organization struggles. Justifying new servers is always a business decision. Translate server features into the clear way to fix those bottlenecks or reliability problems. How will this hardware expenditure save the business money, protect it, or allow it to generate more revenue? If you cannot answer these questions, no amount of technical assessment will do.
For app's sake
Are these new server features necessary to support important next-generation business tasks and upgrades? That is where the justification for a server refresh matters -- if a new application will succeed or fail (slow = failure) due to its hardware.
If you're upgrading applications, you probably need to meet radically different system requirements. Consider a server to host Microsoft Exchange. Exchange Server 2013 requires an x64 processor (often with multiple cores) and 8 GB of memory. That's twice as much as Exchange Server 2007, and Exchange Server 2003 ran on just a 733 MHz 32-bit processor and 512 MB of memory.
New server hardware also takes virtualization software to a new level. Older servers might only support several VMs per host. The introduction of Intel-VT and AMD-V command sets in newer processors enables vastly better performance under hypervisors and allows high levels of consolidation.
New features aren't all about muscle. Memory-channel storage brings nonvolatile flash memory to DIMMs, opening up vast data storage without ever leaving the memory module. This enhances VM performance and reliability. For scientific or big data tasks, graphics processor unit integration takes the burden off the main processors.
Technically, we need this
New servers address business directives as well, such as saving energy. The amount of energy used to run and cool data center servers outstrips demand from nearly all other parts of the business. Next-generation servers use efficient processors and support the same amount of workloads on fewer physical hosts. If the company is undertaking an energy use reduction project, the up-front cost of new servers should be considered a necessary investment.
Another example is regulatory compliance directives. A business intent on strengthening its security posture to meet compliance obligations might invest in servers that incorporate a Trusted Platform Module, Intel Trusted Execution Technology or native file encryption. The up-front cost of more secure servers offsets the potential cost of fines -- or worse, damage from a security breach.
We don't need the latest and greatest
Even if applications aren't straining your current hardware's capabilities, you might budget for new servers to eliminate high maintenance and support contract costs, and the expense and time involved in tracking down replacement or upgrade parts. As vendors shift attention to new system designs, it puts pressure on IT shops to upgrade -- or get left behind.
Slow, ineffective support's intangible effects on application uptime and system reliability factor into refresh decisions. In a virtualized data center, old, unsupported hardware threatens all applications that use that one host. Imagine the chaos when a server with 10 VMs goes down without immediate vendor support, field service or replacement parts available.
This is a simple argument of risk vs. cost avoidance. The cost of renewing support and maintenance agreements past the initial coverage period is almost always unfavorable. Compare the cost of renewed or third-party service agreements to the cost of new servers. To avoid busting the budget, organizations can acquire new servers for critical applications first and then relegate older servers to handle basic business workloads in clusters that mitigate downtime from hardware faults.
New servers also address data center capacity planning pressures, wherein internal or customer growth demands more storage, bandwidth, processor cycles and memory space. New servers offer significantly greater available computing resources.
Don't refresh the servers just because it's on your capacity planning roadmap. Service outages, poor customer experience or internal complaints about lost productivity are factors that warrant capital expenditures on servers.
Use that purchasing power
Most IT organizations buy servers based on their existing infrastructure, replacing the old HP ProLiant fleet with the latest generation ProLiant servers, for example, and so on. But if your server vendor doesn't offer all the features or capabilities next-generation workloads require, a refresh is the perfect time to change companies.
In other cases, server design changes present interoperability problems in heterogeneous environments. Going with a different model from the same vendor often resolves this issue.