IBM Systems Magazine, Mainframe - November/December 2013 - (Page 32)

TECH CORNER BEYOND One Dimensional Alternative metrics deliver a more complete view of hardware for server RFPs B usinesses and government agencies frequently use request for price (RFP) documents to bid out work and capital acquisitions. They're the rule in government processes, but are also used in the private sector in competitive situations and when transparency and fairness are enforced by business practices or regulatory scrutiny. When the bid involves the purchase of servers and its scope is such that multiple vendors have competing products with different machine architecture, two key questions arise: 1 2 How much capacity is the bid for? How can that be specified in a way that allows all vendors to compete? The problem is these metrics lack sufficient information to position machines as qualified or unqualified for the work at hand. They either ignore or minimize the contribution of resources not related to thread count and thread speed, which leads to defining relative capacity from a one-dimensional view of a solution that's at least Common Metrics A common metric is a standard adopted by the user to express the size and performance of the desired machine. Clients have adopted such industry standard benchmarks as Transaction Processing Performance Council's TPC-C, SAP Standard Application Benchmarks, and Standard Performance Evaluation Corporation's SPECintRATE, SPECjbbRATE, SPECjappRATE and SPECfpRATE. Before the introduction of Nehalem to use Processors x MHz, only bids for Intel* were commonplace. Now, composite metrics are available, including Gartner's Relative server Performance Estimate 2 (RPE2). 32 // NOVEMBER/DECEMBER 2013 two-dimensional. At best, they specify a single point for the other resources but cannot adjust that point to match the workload. Getting an idea of what the metrics reveal requires plotting the values of RPE2 versus the throughput metric called TP, which is defined as thread count times thread speed. Thread speed is taken as clock rate times Figure 1: Relative Server Performance Estimate 2 RPE2 Joe Temple is a retired IBM Distinguished Engineer and principle consultant of Low Country North Shore Consulting. 0 20 40 60 80 TP - Adjusted GHz 2.9 GHz SB POWER7 2.6 GHS SB POWER7+ 100 120

Table of Contents for the Digital Edition of IBM Systems Magazine, Mainframe - November/December 2013

Table of Contents
Editor's Desk: On the road with analytics
IBM Perspective: Extending IBM's commitment to big data and analytics
Trends: IBM Capacity Management Analytics for zEnterprise cost-effectively ensures optimal performance
IT Today: IBM DB2 Analytics Accelerator V4.1 expands the value of high-performance analytics
Case Study: The State of Consolidation: Oklahoma improves service levels by consolidating its IT on zEC12
Cover Story: Big Data's Big Impact: zEnterprise hybrid technology helps redefine business analytics
Feature: The System z Crystal Ball?: Teach the mainframe to predict the future using analytics and automation
Tech Corner: Alternative metrics deliver a more complete view of hardware for server RFPs
Administrator: QMF extends its analytics capabilities to help meet the demands of big data and mobile
Solutions: TestBase; Vanguard Offline
Stop Run: Historic property has much to teach retired IBMer
Reference Point - Global Events, Education, Resources for Mainframe
2014 Mainframe Solutions Edition Product Index

IBM Systems Magazine, Mainframe - November/December 2013