IDC Says Power Systems Trumps X86 In Resiliency
August 13, 2012 Alex Woodie
Productivity losses due to unexpected downtime were six times higher in X86 environments than Power Systems environments says a study by IDC. The study, which was commissioned by IBM and turned into a white paper that was recently posted to the IBM website, found that the cost of IT staff responses to downtime was 10 times more in the X86 environments than it was in IBM Power Systems environments.
In recent months, IBM has posted a slew of new white papers on its website that highlight the server platform and its competitive advantages. In addition to the 13-page white paper by IDC analysts Jean Bozman and Randy Perry, titled IBM Power Systems and Their Support for Business Resilience in Challenging IT Environments, IBM has posted a new IBM i roadmap white paper (which we covered last week) and a new white paper touting PowerVM’s capabilities against its competitors (which we’ll cover in the coming weeks).
Bozman’s and Perry’s white paper, which can be downloaded at the systems/i homepage at www-03.ibm.com/systems/i, highlights the need for highly reliable and available computer systems to service today’s 24x7x365 business environment. The analysts write about some of the well-known reliability, availability, and serviceability (RAS) features of the Power Systems architecture, such as the error correction, redundant I/O, and hot-swappable power supplies and disk controllers, and some of the less well-known parts, such the redundant and hot-swappable power regulators, a feature that’s only found on the biggest Power 795 servers.
The most interesting part of the paper lays in the comparisons between X86 and Power Systems servers. To arrive at the figures mentioned at the beginning of the story, the analysts borrowed data collected over the last four years from interviews it did with 17 organizations that had migrated workloads from X86 to a Linux-based Power Systems server (yes, this is a Linux comparison). Seventeen cases hardly make this statistically significant in a scientific sense, but as my 4-year-old recently learned in preschool (and is fond of repeating about the house), “You get what you get, and you don’t throw a fit.”
The IDC researchers found that, in these 17 organizations surveyed from 2008 to 2011, unexpected downtime decreased from an average of 11.46 hours per year to 1.88 hours per year for the servers running the applications at issue. Annual end-user productivity losses due to downtime were $6,718 per 100 users on the applications running on X86.
After the apps were moved from the X86 machine to the Linux-based Power Systems machine, the annual productivity loss due to downtime was $1,099 per 100 users. The cost of IT staff responses to downtime (based on the total salary of the IT professionals who are tasked with fixing the outage) was $1,113 per year for the X86 apps, versus $89 for the Power Systems apps.
Overall, organizations that moved workloads from X86 servers to Power Systems servers were able to reduce unplanned downtime by an average of 84 percent, the IDC analysts write. They continue: “The combination of reducing the total number of x86 server platforms to be managed and migrating workloads to the Power platform reduced downtime hours from as much as 120 hours per year to an average of 1.9 hours per year per site, including some customer sites in which downtime was virtually eliminated.”
These numbers jibe with other studies of downtime by platform. Laura DiDio, the former Yankee Group analyst who now shares her IT industry insight at her Information Technology Intelligence Corp, has given the AIX-based Power Systems platform the top spot in reliability since she began tracking the number of downtime incidents per platform at ITIC in 2009. AIX had the fewest number of downtime incidents among 19 different platforms for the 2011 survey, which ITIC released in January 2011.
However, the 2011 ITIC survey also found that Windows Server 2008 has made a surprising improvement in resilience from 2009 and 2011. All the platforms showed improvement, the group said. Most of the major downtime incidents for all platforms were due to integration issues, such as patches gone awry in highly customized environments, and not problems in the server hardware or OS.
IDC also studied how downtime means different things to different companies and industries. The analysts cite a separate, larger IDC survey that found nearly half of organizations would consider a one-hour outage to be a “business disaster” or have a “severe business impact.” More than half, however, say a one-hour outage would have a moderate or small business impact, or have no effect whatsoever.
Different industries have different approaches to downtime, which is completely logical. The IDC found that, if you work in the financial services industry, downtime will cost you an average of $10 million of revenue for every hour your systems are down, while employee productivity is hurt to the tune of $3.6 million per hour, the IDC said. This is why the financial services industry spends the most on building high availability into systems and why it has the lowest amount of downtime: an average of 3.57 hours per year, according to the IDC.
The manufacturing and retail industries have similar levels of downtime, an average of 8.01 and 7.75 hours per year, respectively. Downtime costs manufacturers about $3,006 per hour in lost employee productivity, while it costs retailers $2,580. However, revenue losses due to downtime are much greater in the retail business ($397,500 per hour) than they are in the manufacturing business ($59,930 per hour), largely due to the fact that retail is a real-time business, while manufacturing is batch oriented.
The healthcare industry, interestingly, had the most amount of downtime per year–an average of more than 21 hours per year–despite the fact that it had the third-highest hourly revenue loss from downtime ($157,500 per hour). The public sector had much less downtime than healthcare organizations–about 9.4 hours per year–and had the lowest productivity loss at $850 per hour, the researchers found.
Companies take different approaches to addressing downtime. IDC says its studies indicate that 59 percent of organizations prefer a hardware-based high availability solution (such as IBM’s PowerHA disk-level clustering) rather than a software- or middleware-layer high availability solution, such as the third-party logical replication products that are widely used in the IBM i market.