Problem solve Get help with specific problems with your technologies, process and projects.

Centralizing IT on Linux mainframes

Linux has revitalized the mainframe. Learn how today's mainframes are reducing TCO and TCU through consolidation, better performance and more. Expert Ken Milberg explains why big iron is back (again) and why deploying Linux is the "logical evolution" of the mainframe.

Linux server consolidation usually involves distributed clusters, but larger, vertical Linux systems may benefit IBM's System p. The midrange platform allows for the operation of either System p supported versions of Linux (RHEL and SLES) or AIX, IBM's Unix distribution. But what about the mainframe? Options and some new features available to those migrating to Linux on mainframes will be discussed in this article.

Open source on the mainframe may sound like an oxymoron, but it's not. Linux on IBM System z is proof of this relationship. Competitors can no longer claim that using the IBM mainframe locks them into using only proprietary operating systems. Linux on the mainframe really benefits from IBM's experience with virtualization and hardware integration. For example, managing multiple workloads using context switching allows users to run literally hundreds of Linux instances on one mainframe. While managing these multiple workloads, the ability of System z to switch between multiple processes at a faster rate than X86 servers clearly improved.

The availability of Java based applications such as WebSphere and strong back-end databases such as DB2 on the mainframe make the platform that much more appealing. From a staffing perspective, it solves a problem that many predicted: the challenge of supporting legacy applications on the mainframe as the baby boomer generation starts to retire. It's ironic that Generation X and Y, who have been trained on Java and Linux, are doing well in this type of mainframe environment; that is, they are the ones taking care of the boomers on their mainframes. So the perception that everyone supporting mainframe operating systems is over 60 is just not based in reality.

That said, it would not be inaccurate to say that there are simply more people learning distributed architecture on the systems side of management, like Unix and Linux, than mainframe technology. IBM has recognized this and has actually started a joint effort with approximately 230 universities worldwide. Their goal is to train up 20,000 mainframe engineers within the next three years. This will be done by developing courses, training professors and providing access to mainframe facilities. IBM has also been working with SHARE to build a global community for young professionals' interests in mainframe computing called the zNextGeneration.

For those that have already migrated to Linux from Windows or Unix servers, moving to Linux on the mainframe is the logical evolution, as part of a large scale consolidation effort towards scaling up and centralization of systems and moving away from the distributed decentralization mode of the late 1980s and early 1990s. The mainframe even offers z/OS Unix Systems Servers (USS), which is a certified Unix implementation (XPG4 Unix 95) optimized for mainframe architecture. Rather than locking you in, today's mainframe is all about choices.

Reasons for migration to Linux on the mainframe
Reliability, Availability and Serviceability (RAS) - RAS refers to features which enable the mainframe to avoid unscheduled downtime. The RAS capability of the mainframe far supersedes what is available on any other platform. The reliability component refers to design features which can actually detect faults on the system, stop processes and report problems rather than continue operations with bad calculations. Availability features refer to designs that allow the system to continue to function even with systems faults. The serviceability part of this equation refers to how easy it is to diagnose a server system that has malfunctioned. Some important features include journaled file systems for file system reports, concurrent updates for system resources, failover capabilities, computer clustering capabilities, automated fault isolation and self-healing.

These features are important to disaster recovery (DR) and business continuity. Ever try performing a DR exercise on distributed systems? The essence of the mainframe lends itself to DR exercises and real DR solutions, while distributed disaster recovery plans are often nothing more than vaporware.

Data center and server consolidation – One cannot emphasize enough the strategic advantage of being able to scale and consolidate vertically as opposed to horizontally. Decentralization of servers has contributed to extremely large horizontally based server farms, which are extremely difficult to manage. This is one important reason why virtually every other project today is based on server consolidation. Distributed server farms don't scale well and lead to all types of availability and performance problems. Today's trend is to consolidate to fewer, faster CPUs, rather than rely on greater numbers of servers or CPUs. This type of consolidation reduces real-estate footprint and ultimately cooling and energy costs. IBM's Hypervisor based virtualization, z/VM, provides the capability of running hundreds of Linux servers in one LPAR. IBM's dynamic LPAR technology also enables one to create up to 60 LPARs on one System z. Furthermore, unlike virtualization solutions available on commodity PC hardware, such as VMWARE and XEN, IBM's virtualization is also tightly integrated with the hardware. Its 40-year virtualization maturity level is also unmatched in terms of anything else we have today. Consolidating on the mainframe with Linux allows for physical consolidation of hundreds or thousands of distributed servers into several partition based mainframes.

Total Cost of Ownership (TCO) and Total Cost per user (TCU) - The cost of a mainframe is obviously much larger than the cost of acquiring a PC. At the same time, it is important to look at all costs and to understand the difference between the Total Cost of Acquisition (TCQ) and Total Cost of Ownership (TCO). For example, TCO includes licensing costs, energy and power costs, staffing costs and floor space.

The Total Cost of technology includes:

  • Hardware - Maintenance, indirect costs such as power and cooling and real estate- datacenter.
  • Software
  • Manpower costs – The costs involved in maintain software and hardware. A recent study by Arcati Research, compared the total cost per user of many different architectural platforms. They projected that by 2010, mainframe costs are approximately 25% of PC server and about 33% of Unix servers.
  • Provisioning – New Linux virtual machines, using z/VM can be configured in minutes rather than the type of time it takes to configure similar solutions on distributed systems.

Resource utilization – While mainframes typically run at close to 100% CPU utilization, distributed systems are close to 20%. While the latter number is increasing due to the advent and improvement of virtualization on PC-based Linux distributions (i.e., Xen implementations on SLES and RHEL), one cannot compare the maturity of these products with the 40-year maturity of virtualization on the IBM mainframe.

Security – Perhaps more than any other architecture, the IBM mainframe is as secure of an environment as you can get. This has always been one of the primary advantages of this environment. The mainframe provides for the highest level of security, a level 5 Evaluation Assurance Level (EAL). Have you ever heard of a mainframe being hacked? Probably not, because it just does not happen. The mainframe has security built into its core. It even has its own networking infrastructure, which certainly adds to its inherent security.

I recently had the opportunity to play around on an IBM mainframe with Linux partitions. Other than the increased performance, the same commands that work on your x86 or System p Linux on Power (LoP), are available.

OS version:
[root@etpglmy /]# cat /etc/redhat-release Red Hat Enterprise Linux Server release 5.1 (Tikanga) [root@etpglmy /]#

Hardware model info:
[root@etpglmy proc]# more cpuinfo vendor_id : IBM/S390 # processors : 1 bogomips per cpu: 3407.87 processor 0: version = FF, identification = 01AD0E, machine = 2094 [root@etpglmy proc]#

Disk space using df:
[root@etpglmy proc]# df Filesystem 1K-blocks Used Available Use% Mounted on /dev/mapper/system-root 2805112 2698000 0 100% / /dev/mapper/system-opt 1161496 230192 871352 21% /opt /dev/mapper/system-var 380888 67224 294004 19% /var /dev/mapper/system-home 218208 16588 190356 9% /home /dev/dasdb1 76616 11820 60844 17% /boot tmpfs 255676 0 255676 0% /dev/shm [root@etpglmy proc]#

Concerned about the market share of the mainframe? Some are hesitant to move to an architecture that they see as outmoded. Actually, the truth is that the mainframe market share is growing in leaps and bounds:

  • IBM System z boasts five consecutive quarters of revenue growth and eight consecutive quarters of MIPS growth. In 2006, mainframe revenue outgrew platforms on Windows, according to IDC, an important driving factor being its ability to run Linux and Java applications. Overall, the mainframe market grew 8% in 2006.
  • In a recent Gartner Group Study, Nationwide Insurance consolidated hundreds of distributed servers into the IBM mainframe (using Novell's SUSE Linux) and projected savings of $15 million over three years with a 50% decrease in hardware and OS costs, along with an 80% decrease in floor space and energy and cooling requirements.

    The advantageous TCO was proved because of the large number of Linux virtual servers that were created from the consolidation of physical X86 PCs. They chose mainframe virtualization because of its maturity, low overhead on most workloads, better application isolation, high availability of hardware and a simpler DR solution.

Mainframes do not have to be purchased through Big Blue, but can be acquired through IBM Business Partners. Also, IBM has just announced the introduction of the z10, which features many new innovations and impressive performance. Far from extinct, the mainframe is back and living through open systems and Linux.

About the author: Ken Milberg has worked for both large and small organizations and has held such diverse positions from CIO to Senior AIX Engineer. Currently, he is the president and managing consultant of Unix-Linux Solutions, a NY based IBM Business Partner. Ken also writes for IBM Systems Magazine, Power Systems edition and for IBM developerWorks. He is also the founder and group leader of the NY Metro POWER-AIX/Linux Users Group.

Dig Deeper on Linux servers

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.