IBM’s Servers Are Running Out of Room for Improvement

Insights from ITIC’s Reliability Survey

For insights into server uptime and reliability, we can refer to the annual survey conducted by Information Technology Intelligence Consulting (ITIC), led by industry analyst Laura DiDio. Since its inception in 2009, this survey has provided valuable data on server performance.

The latest 2023 Global Server Hardware, Server OS Reliability Survey reveals impressive statistics. According to the survey, 88% of users with IBM’s Power10 server—running IBM i, AIX, or Linux—reported an uptime of 99.999999%. This means they only experienced about 315 milliseconds of unplanned downtime each year due to system flaws or component failures.

Even more impressive is the IBM z16 mainframe, which achieved 99.9999999% uptime. In this case, 96% of its users reported just 31.56 milliseconds of downtime annually. This outstanding reliability makes the z16 the most dependable system available today.

Comparison with Industry Peers

In contrast, the rest of the server industry did not perform as well. The Lenovo ThinkSystem, previously known as the IBM System x server before its sale to Lenovo in 2014, ranked third with an average of 31.5 seconds of annual downtime, equating to six nines of availability.

According to the ITIC survey, other reliable systems included those from Cisco, Hewlett Packard Enterprise, and Huawei, which reported between 1.27 and 1.39 minutes of downtime per year. The Fujitsu Primergy, with 5.9 minutes of downtime, was the only server in the five nines club.

The reliability figures dropped considerably from there. For example, Dell PowerEdge servers had 24 minutes of downtime annually, while Oracle's systems reported even higher downtime levels.

Comparing these statistics to past reports shows a gradual improvement in IBM’s server reliability over the years. In 2021, ITIC reported that Power servers averaged 1.49 minutes of downtime per year, while ITIC's earlier reports indicated that 61% of IBM Power Systems achieved 99.999% availability in 2016.

To summarize, over the past eight years, IBM i customers have seen unplanned downtime decrease from over five minutes to less than one-third of a second annually. Meanwhile, System z mainframe users experience an impressive 31.5 milliseconds of downtime.

Challenges and Future Improvements

IBM deserves praise for its continued efforts to enhance the reliability of its Power and System z servers. The company is well-regarded for developing systems with superior reliability and security compared to competitors.

However, it’s important to acknowledge that further improvements may be challenging. The current levels of unplanned downtime are already so low that achieving even smaller increments may have little practical impact on users.

As noted by IT Jungle, many companies report that downtime costs them over $300,000 annually, with 20% stating it exceeds $5 million. This shows that there remains significant room for improvement across the industry.

For IBM, the situation is different. With Power10 servers experiencing less than one second of downtime annually, many businesses face minimal losses. DiDio reports that the average Power10 enterprise spends only $7.18 per server each year on remediation due to unplanned outages. Mainframe users spend even less.

While aiming for ten nines of reliability in future System z mainframes or nine nines in upcoming Power11 servers sounds impressive, these improvements might not make a big difference for customers. 

However, IBM can still improve its systems in other ways. Security threats, which can cause unexpected downtime, need constant attention. Issues with applications or data can also lead to downtime, but these problems are often outside of IBM's control. 

Right now, IBM is leading the way in building reliable servers, combining processors, RAM, drives, power supplies, network adapters, and firmware. This is a major achievement for the industry and deserves recognition.