This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
1. - Is it time for those servers to go?: Read more in this section
- When do servers go past their prime?
- Ask the right questions to determine server upgrade strategy
- Five tips for an effective server refresh
Explore other sections in this guide:
The server lifecycle of a data center may vary, and understanding the right reasons to upgrade can help you figure out when to put aging servers out to pasture.
Businesses rely on servers that provide the computing power for critical applications, allow users and customers to interact, produce work, and generate valued revenue. As with most business assets, servers have a limited working life and need to be replaced periodically. Eventually, every server comes off lease or reaches end-of-life; new features and capabilities are constantly appearing, promoting the acquisition of new servers. Yet, technologies like virtualization are extending the working life of modern servers. As a consequence, the actual point where a new server is justified has become a bit murky. Let's cover the top five reasons to consider a server upgrade.
Upgrade servers that do not provide important features
Each new server model introduces support for the latest features, such as new memory types, processors with new extensions, more aggressive power throttling or conservation features, or advanced thermal management such as liquid cooling. TechTarget's 2012 Data Center Decisions survey reported that 36% of IT professionals replace older servers to improve the data center's energy efficiency. As another example, it would be almost impossible to achieve worthwhile levels of virtualization on systems with processors that did not include Intel VT or AMD V extensions. New server acquisitions are often required when new data center initiatives -- such as consolidation projects -- are initiated.
Upgrade servers that do not provide satisfactory workload performance
Servers do not wear out in the traditional mechanical sense, but workload performance will typically degrade over time as more demanding application updates and patches converge with a growing user base. For example, enormous storage traffic demanded by the current user/customer base may simply overwhelm the server's network bandwidth, or a current software release may rely on chipset features that don't exist on the current system -- both resulting in poor response times and performance problems. If a host server can no longer meet the growing computing demands of applications and users, newer and more powerful systems may be required.
Upgrade servers that have become unreliable or cannot be maintained
Most organizations do not service or repair their own server hardware. Instead, they rely on annual maintenance agreements between the business and the server vendors (or even third-party vendors). As years go by, parts become scarce and vendors often focus their service organization on more recent product offerings, charging far more money to service older systems. Eventually, the cost of maintenance contracts becomes prohibitive and an upgrade to a later system is easier to justify in lower service agreements and troubleshooting/maintenance costs.
Upgrade servers that are out of computing resources
A business may dictate a certain minimum of workloads allocated to each physical server to ensure satisfactory consolidation. In other cases, IT may establish a minimum pool of computing resources to be maintained for workload-balancing purposes. It may be possible to optimize the computing resources allocated to any underperforming workloads, or even perform workload balancing to move some workloads to other available servers (freeing resources for the remaining applications). But older servers that fall below these minimum capabilities are prime candidates for upgrades.
The server lifecycle must support management initiatives
Companies with larger data centers often standardize on hardware-based management subsystems such as Dell Remote Access Card (DRAC) or Integrated Lights-Out (iLO), or may opt for systems management tools that can examine and control the hardware infrastructure. While this works well in homogeneous environments, businesses with a mix of heterogeneous systems can face serious gaps in management coverage, especially in older servers that are improperly or incompletely supported. Older systems are frequently replaced to provide a full picture of data center status and ensure the most granular management control.
The end of the server lifecycle is inevitable, but the timing and urgency of those upgrades can vary dramatically depending on the size and needs of your data center. Still, IT professionals should recognize some common situations that may warrant a server upgrade cycle: typically any situation that impairs the servers' capabilities, compatibility, reliability and control.