Many organizations, when faced with budget challenges, put off capital expenditures (capex) and seek alternatives to acquiring new hardware platforms, such as lengthening server life cycles and extending software licenses. This pattern of stretching the useful life cycle of servers has a number of near-term benefits for customers in terms of depreciating assets over a long period of time or prolonging an existing lease. But if a transition to new technologies has been deferred too long, then the time comes when the system has fallen far behind the performance and cost-efficiency levels being offered by multiple vendors in the marketplace today. This has been especially true in recent years, as the performance of processors has more than doubled each year, based on the emergence of multicore, multiprocessor system designs with improved system speeds.
This paper demonstrates that a buy-and-hold strategy can actually add costs to the datacenter, for a number of reasons, as systems age in place:
Hardware maintenance costs rise over time, and performance lags behind more current server offerings.
Energy efficiency is not as advanced in older server models — leading to rising power/cooling costs in the latter years of the server's usable life cycle.
Applications software and systems software fall behind the current versions
available in the marketplace, and security may require frequent updates. After
five years of use, the cost of replacement climbs.
To understand more about server replacement cycles and ongoing operational costs, IDC studied many sites that remained on an existing platform long after its initial introduction and then upgraded. When comparisons to the succeeding generations of technology are made, customers have found that the increase in scalability and performance of newer systems, combined with a reduction in server "footprint" size and overall power/cooling requirements, resulted in significant reductions in ongoing costs, or opex, per 100 end users supported. This paper describes HP ProLiant Gen8 servers, the technology on which they are based, and the way they address many of the causes of operational costs found at customer sites. The hardware and software capabilities of the HP servers were designed
to be proactive, reducing the effort and knowledge required to run the server systems and leveraging automation to reduce maintenance costs and IT staff costs.
The Cost of Retaining Aging IT Infrastructure
Saving Money by Leveraging Technology
Refresh: How Current Accounting Misses
Most organizations continue to purchase their servers and IT equipment and then, following the initial investment, use a "standard" financially derived amortization period — often five years. Typically, this has led to a useful server life cycle of three to seven years, depending on the type of platform, operating system, and workloads being used. While most IT shops replace their x86 systems every three to five years, they have tended to hold onto their Unix servers supporting mission-critical workloads for longer periods of time — generally five to seven years or more — given the importance of the workloads being supported.
After acquiring and capitalizing equipment and then initiating the amortization period, most IT managers avoid making further changes, resulting in a delay in updating the systems or providing a technology refresh. Often, they do not replace the equipment before its normal depreciation cycle runs its course, as long as the system is performing adequately and meeting availability requirements. This approach to server replacement/renewal cycles misses an important assessment of the actual conditions and cost factors experienced. Instead, it relies on the calendar to determine when the server should be replaced or refreshed with new technology.
During this time, system administrators may work to repeatedly upgrade and reconfigure servers in support of workloads rather than to consider a fully burdened cost assessment highlighting the cost reductions that could be gained by replacing the servers sooner. In many cases, a cycle of repeated upgrades, security patches, and rising maintenance and management costs can accelerate, over time, if the life cycle of the server is extended to four years or more.
Current Capital Constraints Contribute to Lengthened Server Life Cycles
IDC's supply-side data for the worldwide server market showed patterns of lengthened server life cycles. The data documented the delay and deferral of many mid-range and high-end servers, starting in fall 2008 — at the onset of the economic downturn — and continuing through 2010. At that point, IDC saw an uptick in midrange and high-end server sales, including non-x86 server systems, that was fed by a wave of technology upgrades across those server classes. At the same time, the level of investments in x86 server technology has grown, now generating more than
95% of server unit shipments per year and more than 65% of server market revenue per year.
Following the 2008–2009 downturn, the recent rise in server market revenue reflected that a technology replacement cycle was under way. There were also indications that IT organizations needed to acquire new hardware platforms while taking advantage of leveraging virtualization to consolidate workloads onto fewer server platforms for the sake of operational efficiency and reduced IT costs.
Focusing on Operational Costs
The drive to reduce capital expenditures is strong and understandably so given the current economic climate. However, IT managers also know that the need to address opex within the datacenter is equally important.
Although IT managers did a good job of capping IT spend on servers and storage throughout the economic downturn, costs on the operational side of the IT organization continued to grow. Starting in the late 2000s, and throughout the
economic downturn period, costs for maintenance and management, along with costs for power/cooling, have soared. Power/cooling costs grew eight times as fast as server acquisition costs — and costs for maintenance/management, viewed as a category, grew four times as fast as server acquisition costs.
Maintenance/Management Costs and Power/Cooling Costs Rise over Time
As Figure 1 shows, by 2010, maintenance/management costs generated twice as much in total IT costs as server acquisition alone — and power/cooling costs grew enough to nearly equal server acquisition costs worldwide; in some cities, power/cooling costs already outstripped the server acquisition costs. Meanwhile, the worldwide installed base now stands at more than 35 million units and is projected to grow even more. Fortunately, the growth in logical servers (virtual servers or virtual machines/VMs) is providing more usable capacity per physical server — and this is
improving server resource utilization for each server — over time. View Chart and Continue reading this White paper here >>