Financial comparison of fixed-time vs. optimal-time (JIT) refresh cycles in a virtualized server environment

Server virtualization fundamentally changes the economics of server refresh cycles by enabling incremental, just-in-time server refresh. The cost savings and technical advantages of dynamic, just-in-time server refresh for enterprise-class production environments are significant, yet most companies continue to adhere to pre-virtualization rules-of-thumb to guide their refresh cycles and operate their servers on a 3-6 year time horizon.

This analysis contrasts two refresh cycle strategies. The first scenario is based upon a common server refresh practice in which servers are operated for a fixed time period and then replaced. The second scenario is based upon replacing servers on a flexible, just-in-time schedule determined by a new capacity planning approach that simulates demand and capacity and finds the least cost refresh strategy, incorporating both hardware and software license costs.  

The results show significant, permanent savings available to companies. By essentially leveraging the exponential processor performance gains available, and optimally translating that into increased compute capacity per software licenses, companies can significantly lower processor counts and software licenses to perform the same workload.      

Both scenarios make the same material assumptions about service demand, target utilization rates of the servers, costs, inflation rates, and discount factors. Both scenarios print an “evolution” of server utilization, server processor counts, and capacities. Capacity measures are determined from standard benchmarks, and are ideal for this type of such relative analysis. Furthermore, both scenarios start with the same initial values for capacity, server quantity, processor and core set sizes, and initial inventory cost. In this analysis, as in real-world practice, it is the relative cost difference of the two strategies, as they evolve over time, which is important.

The first scenario is based upon a fixed-replacement time and assumes the initial server inventory is not replaced for four years. This is within the parameters of the typical operating period, of 3-6 years, which many companies adopt for server “useful” life.

Next >>