Editor’s note: This is the first in a series of four articles on server hardware equipment refreshes. In part one, you’ll find strategies for getting the timing right for trading in old hardware for new -- and when to extend a server’s life instead.
As servers age, IT shops face inevitable questions about when to upgrade their hardware. Organisations wrestle with a number of issues on their way to making a decision, such as expired warranties, the inability of aging hardware to support growth, unsupported components and firmware, and a lack of support for the latest OSes. Managing legacy hardware that is out of step with new software can easily become an albatross around an IT manager’s neck.
But, in many cases, ripping and replacing existing hardware isn’t easy either. With new purchases, you face a litany of considerations: hardware and processor compatibility, virtualisation requirements, power and cooling, and new form factors (such as a move from rackmount to blade servers) -- not to mention new business-side services.
As you consider whether to make these potentially costly purchases, answering the following questions can help you decide: Is a new server purchase more cost effective than extending server life and the expenses that incurs? Would the cost of new hardware be worth preventing a server breakdown?
Ultimately, you have to weigh the costs and benefits of purchasing new servers or extending existing server life.
Factors that prompt a server hardware refresh
1. Hardware compatibility demands. For aging hardware, a major reason for making a decision to upgrade or not is compatibility.
When a business wants to upgrade applications or core services such as databases, its existing hardware may be obsolete. Although compatibility is more of a factor for proprietary chipset-based servers, where the complete hardware and software stack is typically interdependent, it can also be an obstacle for independent software vendors where the software runs on x86 environments. If you do not use a validated server model as listed on a hardware compatibility list (HCL) or if existing servers don’t meet hardware specifications, your only option is to refresh with new server-supported hardware.
When upgrading to newer-generation OSes, compatibility is also a factor. Hardware vendors typically phase out legacy driver hardware support within new OSes. For that reason, existing hardware eventually lacks official OS driver support.
Finally, existing servers may support only 32-bit architecture, which has only 4 GB of physical RAM. This limited RAM may not meet business requirements, and the lack of capacity for an upgrade may necessitate a server hardware refresh and OS (or both) to 64-bit server architecture to support larger memory demands of applications and services.
2. Virtualisation improvements. Your shop may need to perform a server refresh to replace older virtualisation hosts that cannot support the consolidation of more resource-intensive virtualised server workloads. Unlike older servers, today’s hardware is designed to sustain greater virtualisation demands, and hypervisors are now designed for CPU chipset-level features. So a refresh of the existing virtualisation platform can reap the benefits of virtualisation.
Such benefits include higher consolidation ratios and consolidation of server workloads that previously resided only on physical servers. Newer technical features, such as hardware-assisted memory management units, improve memory allocation, and I/O memory management features directly present PCI devices to virtual machines (VMs) to facilitate real-time, network-intensive workload needs.
Still, before you leap into a hardware upgrade, consider the downsides in terms of virtualisation. Newer servers that include new enablement features will likely impose much greater capital cost expenditure compared with the original investment. Managers should also note that upgrading hardware can have ripple effects on virtualisation platforms. In some cases, new hardware may prompt a hypervisor upgrade to reap the underlying chipset benefits.
3. Blade servers or unified computing. Certain organisations may decide to replace existing rackmount servers with blade server platforms. Adopters of blade technology will likely achieve a much smaller consolidated data centre footprint. Rackmount servers inherently have a much higher overall rackmount space and back-end port connectivity requirement. In comparison, using blades within a shared backplane chassis, individual blade servers will use this shared connectivity to reduce overall SAN, LAN and power requirements. And, blades provide heavy reductions of data centre real estate consumption.
But organisations should choose blades over rackmount servers only when they are confident that these servers have matured beyond the teething issues of prior generations. A refresh may also be influenced by emerging unified computing technology platforms such as Cisco Systems’ Unified Computing System and Hewlett-Packard’s BladeSystem Matrix, whose underlying workload is blades.
4. Company growth. A business may also turn to a server refresh because existing servers hamper business growth. Several business-related factors make server performance levels even more important.
New requirements may include increased demand for new business from existing customers, acquisition of a new company, or a new revenue-generating strategy that requires additional application support, such as a new customer relationship management system.
Business growth requires infrastructure to facilitate these demands, which involves continuous risk mitigation by performing regular capacity analysis to review existing server infrastructure spare workload capacity.
If expansion capacity isn’t available, regular reporting may provide enough evidence to warrant a refresh of existing server hardware onto more capable platforms. Risk reviews also assess whether existing servers can meet service-level agreements. If they can’t, a new hardware investment would far outweigh the potential costs caused by the loss of business.
5. Technology roadmap. Consider the broader technology roadmap to ensure that server refresh cycles align with other refresh plans. Your networking infrastructure, for example, may also need a refresh to newer converged or Fibre Channel over Ethernet environments, or your business may be planning to upgrade existing business applications that require a newer server platform.
Daniel Eason is a UK-based infrastructure architect at a multinational company. He also has a personal technology blog: http://www.vmlover.com.