Are Some RISC-Based Clusters Easier to Manage Than Others?

In February 2004, TechWise Research published a paper entitled: Total Cost of Ownership for Entry-Level and Mid-Range Clusters. That paper showed that HP OpenVMS/AlphaServer clusters cost, on average, $248,000 less to manage over a three-year timeframe than IBM AIX/pSeries clusters. TechWise Research decided to conduct a follow-up study to probe specifically into the area of cluster management costs. The purpose was to better understand the differences between HP and IBM cluster management costs, and the possible reasons why HP has an advantage over IBM in this area. The results from this follow-up research are summarized in this paper. For this study, TechWise collected and/or analyzed data from four sources. First, TechWise performed a detailed comparison between the profiles of HP and IBM respondents from the February 2004 study. The purpose was to determine if differences in management costs could be attributed to a sampling bias. HP and IBM respondents were compared on a total of seven profiling criteria. Each of which will be covered in this paper. Second, management and availability data from the February 2004 study were examined in more detail-specifically, the six different costs associated with managing a server cluster (time and money spent managing its nodes, storage, physical environment, operating system/cluster software, applications, and network-related activities). The availability data (based on cluster downtime) was also reexamined because availability directly impacts time spent managing a cluster. Each time a cluster crashes, the cluster team needs to identify the problem, fix it and then re-boot the cluster. Third, we conducted follow-up executive telephone interviews with HP and IBM respondents from the February 2004 study. These in-depth discussions provided further insight into management and availability differences between HP and IBM clusters. Quotes and comments from respondents are included in this paper. Fourth, in addition to speaking with study respondents, TechWise conducted additional research in the area of management to understand the differences between the two brands. One factor that influences management costs is the number of patches that need to be installed to keep a cluster up-to-date. Security patches influence management costs because they take time to install. In addition, if installed incorrectly, a patch could cause system instability, or even worse, a crash. The CERT® Coordination Center is an organization that tracks security vulnerabilities and threats. TechWise performed an analysis of all CERT/CC advisories posted between January 1, 2000 and December 31, 2003, to compare the number of IBM/AIX …