You won't go anywhere fast with critical workloads on distributed Linux servers, so stick to the mainframe to go...
Over 90% of the Linux OSes installed in the marketplace run on x86 architecture, but Linux runs on several other processors. Mainframes offer superior economics over distributed server environments through tremendously high resource utilization rates inherent in the mainframe system design.
Today's Linux kernel is enterprise-strength, with rich reliability, availability and security features, and it is fully capable of running mission-critical applications. IBM reports that 37% -- and counting -- of its largest mainframe customers deploy Linux workloads, eating up spare capacity and preventing new capital expenditures.
If workloads require the highest performance, reliability, availability and security, evaluate running them on a mainframe Linux processor: IBM's Integrated Facility for Linux (IFL) is a lower-cost mainframe processor option just for Linux workloads.
Why deploy Linux on the mainframe?
Mainframe Linux processors execute workloads more quickly and more efficiently than those in x86-based servers. IBM System z processors hit 5.5 GHz, while Intel-based processors in x86 servers, such as the latest Intel Xeon E3, run between 1.1 GHz and 3.6 GHz.
More to consider
Potential pitfalls of Linux on mainframe vs. servers
Mainframes use virtualization and provisioning techniques to share resources and drive up utilization rates. Physical distributed systems share nothing, resulting in very low x86 server utilization rates (in the 5% to 10% range) as compared with 100% sustained utilization rates on mainframes. Even as distributed systems embrace virtualization and provisioning techniques, it is still rare to find an x86 server that runs at more than 40% utilization or a RISC server that exceeds 70% utilization. For example, x86-based servers only hit 40% utilization in a study performed by Solitaire Interglobal Ltd. for IBM in 2012. As long as mainframes are kept extremely busy, they continue to offer the highest return on investment.
Software license fees are usually tied to processor cores. As a result, it can cost significantly less to run Linux workloads on a System z mainframe than on distributed servers. For example, a 2011 Clabby Analytics research report shows it costs over a million dollars less to run a sample workload on Linux on z as compared to a similarly powered x86 configuration.
A System z mainframe can run hundreds to thousands of Linux images on multiple IFL processors. These centralized Linux mainframes are easier to secure than hundreds of distributed Linux servers, where the administrator must secure each network interface on each box and lock down all access points.
Mainframes have the highest security rating in the industry, with an Evaluation Assurance Level of 5+. The mean time between mainframe failures is measured in decades.
About the author:
Joe Clabby is the president of Clabby Analytics, with more than 30 years of experience in the IT industry in marketing, research and analysis. He is an expert in application reengineering services, systems and storage design, data center infrastructure and integrated service management, as well as virtualization and cloud computing.