Monday, September 25, 2017

Table of Contents


Executive Summary

The History of TCO

Background on This Paper

TechWise Research's TCO Model

Configurations Analyzed

TCO Findings Excluding Downtime

The Value of an Hour

Possible Reasons for Crashes and Downtime

Cluster Reliability Findings

Reliability-Adjusted TCOTM for Configuration #1

Reliability-Adjusted TCOTM for Configuration #2

Reliability-Adjusted TCOTM for Configuration #3

Reliability-Adjusted TCOTM for Configuration #4

Differences between Short and
Long-Term Evaluations

Conclusion

Quantifying the Total Cost of Ownership for Entry-Level and Mid-Range Server Clusters

A Detailed Analysis of the Total Cost of Ownership of HP OpenVMS, IBM AIX and Sun Solaris server clusters. (June 2007)

The overall concept of Total Cost of Ownership for server clusters has been an important issue within the IT community for many years. Numerous studies and customer experiences have proven that purchase price alone is not an adequate measurement to compare server clusters from various vendors. Usually other factors, including the costs to manage and maintain the servers, as well as downtime, have a greater financial impact on an organization than just the system's purchase price. Recognizing these factors, TechWise Research developed an analytical approach in 1999 called Reliability-Adjusted Total Cost of OwnershipTM that incorporates management costs and application availability in the TCO analysis. As part of this analysis, server clusters from different manufacturers are compared in terms of the actual number of downtime hours per year that customers typically experience. These downtime findings are then converted into a monetary measurement of the cost differences between clusters which TechWise Research refers to as the "Availability AdvantageTM."

This study focuses on the following three brands of entry-level and mid-range server clusters: HP OpenVMS Integrity, IBM AIX System p5, and Sun Solaris Sun-Fire clusters. TechWise Research last studied these clusters in 2004. Since that time, HP, IBM and Sun have all introduced new server hardware as well as updated operating system and cluster software. These changes all have a direct impact on the Total Cost of Ownership for these clusters. TechWise Research conducted a new study, as a result, to provide more current information on this important topic.

This study involves data from customers as well as publicly available pricing information. TechWise Research surveyed customers that have one of the three types of server clusters installed at their site. The purpose of these interviews was to collect data on the operational costs associated with installing, managing, and maintaining their clusters. Information was also collected on the number of hours and associated costs for various downtime events each company experienced over a twelve-month period. All of these operational data were then combined with current system and service pricing (from IDEAS International) as well as energy costs to calculate the Reliability-Adjusted TCOTM - an analytical approach that factors downtime costs/rates into the TCO analysis. TechWise Research included four main cost components in the TCO analysis. These are the costs to (1) buy the servers and service contract, (2) install and configure the cluster, (3) manage and maintain the cluster over three years (including energy and cooling costs), and, (4) the costs associated with application downtime over three years. Four different cluster configurations from the manufacturers were analyzed. For each configuration, the Reliability-Adjusted TCOTM was calculated at various downtime costs to allow readers the ability to compare the different platforms at a downtime cost rate that most applies for their firm.

Study Results Five potential causes of downtime were used in the analyses in this paper. When all five causes of downtime are considered, HP OpenVMS Integrity clusters averaged the fewest total number of annual downtime hours (1.96), followed by IBM AIX System p5 clusters (6.21), then by Sun Solaris Sun-Fire clusters (9.53). In terms of availability, the HP, IBM, and Sun clusters are 99.978%, 99.929%, and 99.891% respectively. These differences may seem trivial, but in fact they are not.

The reliability findings were further analyzed to determine hardware and software reliability. Isolating crashes caused by the server hardware itself showed that there is little difference in reliability between HP Integrity and IBM System p5 server hardware. Sun Sun-Fire servers, however, were found to have more than double the downtime of HP or IBM. Isolating crashes caused by operating system and clustering software showed clear differences between the brands. HP OpenVMS averages only 0.19 hours of downtime due to the operating system and clustering software. This works out to 99.998% availability. IBM's AIX and HACMP average 4.36 hours (99.950% availability) while Sun's Solaris and Sun Cluster average 4.97 hours (99.943% availability).

Respondents reported that when all costs associated with downtime were considered, including lost sales, lost production and productivity, the average cost per hour of downtime is $145,000. Nearly half of the study's respondents report that their company loses as least $10,000 for each hour their cluster is down. The cost of having a cluster down a few extra hours per year, therefore, can be quite significant. This is why for most companies, management and downtime costs are significant contributors to the TCO for entry-level and mid-range clusters.

The four different cluster configurations analyzed range in price from $100,000 to over $1 million. For all four configurations studied, HP OpenVMS Integrity clusters offer the lowest TCO regardless of the cost associated with an hour of downtime.This represents a change from past studies where Sun clusters used to offer the lowest TCO when there are no costs associated with downtime. Over the past two years HP has been transitioning its OpenVMS business from AlphaServer to Integrity. One reason for this shift was that Integrity servers offered lower cost solutions thanks to economies of scale. This transition has helped to vault HP to be the clear leader in TCO for entry-level and mid-range servers. In all but Cluster Configuration #4, Sun Solaris Sun-Fire clusters offer lower TCO than IBM AIX System p5 clusters at low downtime hourly rates. Once the cost per hour of downtime exceeds a certain level, IBM's better reliability compared to Sun gives IBM a lower TCO than Sun. Cluster Configuration #4 represents the largest systems studied in this paper. These clusters, which are one step below enterprise-class clusters, have price tags over $1 million. In this configuration, Sun clusters have the highest TCO of the three brands regardless of the cost per hour of downtime.

This study showed that the differences in TCO between these three cluster brands can be quite significant and can easily amount to hundreds of thousands of dollars over three years. Applying typical hourly rates for downtime costs, HP OpenVMS Integrity cluster's cost between $290,000 and $1.5 million less than IBM AIX System p5 clusters over a three-year period. Similarly, HP's TCO advantage over Sun ranges from $254,000 to $2.8 million over three years.


To receive a free copy of the full report, please provide the following information.  We will then send you an email with instructions how to obtain your report.







To prove you are not a bot, type the characters you see in this picture.
 
CAPTCHA image

Request Paper     

 

© 1998-2014 TechWise Research, Inc. All rights reserved.
Page Last Updated: 07/19/2014