All too often, I see businesses make server selection decisions based solely on the price of the box and the available budget right now. However, server selection is not a trivial process and the consequences and benefits of choosing servers with business benefits of total cost of ownership, service level agreements, application response time and IT continuity in mind are critical strategic decisions facing every organization.
Over the years, the Intel IT organization has learned and refined our server selection process for our mission critical ERP (enterprise resource planning) applications around the factors below. We have chosen a de-centralized deployment strategy for this environment and select servers based on achieving the benefits below.
• Lower server acquisition costs
• Flexible scalability
• Faster development of new ERP capabilities
• Reduced support costs through standardization
• High availability through clustering
What criteria does your organization use to select servers?
What benefits are most important in the process?
(share your thoughts with the community below)
On Wed September 15, 2010, Sudip Chahal, Principal Engineer, and ERP Infrastructure Architect, Karl Mailman, will discuss Intel IT's approach to server sizing in this free webinar hosted by the SAPinsiders program. We hope you join us.
Learn more by exploring the IT@Intel best practices whitepaper on this topic titled "Sizing Server Platforms to Meet ERP Requirements"
Hi Chris. TCO (Total Cost of ownership) calculations are common in helping to select servers. I understand how server cost and operational expenses can be factored into that equation. Do factors like scalability and faster development fit into a TCO type analysis, or is it treated as a distinct component of the selection?
@ Winston. Yes I believe scalability and agility have a play in TCO. Agility is more of a soft cost variable and may be difficult to quantify in a hard $ calculation. However, in this paper http://download.intel.com/it/pdf/Quantitative_approach_ERP_Server_Sizing_final.pdf, Intel IT does evaluate and estimate the costs of scalability - specifically choosing incorrectly.
The approach here is that if you deploy a server with an intended project life, say 4 years, but find before the end of that 4 year period you need to upgrade the server or resize the application environment, you are placing a significant out of cycle cost onto your business. You may not reach your full deployment life for a variety of reasons including workload peaks (vs average), year-over-year growth rates or requirment for workload failover. The paper shows the different mid-life refresh options for an IT organization and what some impacts to TCO might be (figure 6 in the Intel IT server sizing paper). Clearly the best strategy is to size the server up front for the workload intended to run on it for the entire life of the project - that is a sizing decision.
For this reason, Intel IT deploys 4-socket or larger servers in about 15% of our office/enterprise server enviroment . The paper shows the typical types of workloads (Figure 7) we deploy on 2-socket vs 4-socket in our ERP - enterprise resource planning. The proper sizing of the server is an Intel IT best practice we have found as we seek to maximize the business value of our IT investments
Hope that helps.