Commentary - Even after ten years of successful mainframe deployments, people and organizations not familiar with mainframe Linux often still see the operating system as suitable only for commodity hardware. While Linux indeed runs beautifully on x86 hardware, it is hardly limited to that platform. If you have ever wondered whether Linux can handle complex and scalable mainframe workloads, read on for the exciting truth about Linux/mainframe synergies.
Now that Linux on System z (the definitive term for mainframe Linux) is a reality, its impact – on server consolidation, data center economics, IT reliability and cloud computing – is large and increasing. Today, Linux on System z and the other z operating systems (z/OS, z/VM, z/VSE, z/TPF) are best friends and partners in the mainframe ecosystem.
How We got here: Linux on System z is not new
For those lacking scorecards, May 17, 2010 was the tenth anniversary of commercial/supported main-frame Linux. Its birth followed a late-1990s major shift in IBM thinking leading to acceptance of an open source operating system running on the company’s crown technology jewel, System/390. Fortunately, a skunkworks proof-of-concept project at IBM’s Boeblingen Lab demonstrated feasibility and viability of the unlikely marriage. IBM decided not to release its own mainframe Linux distribution but invited existing suppliers to join the party.
SuSE (later SUSE LINUX AG, acquired by Novell in 2004) quickly offered its prototype implementation, in exchange for access to detailed architectural information supporting ongoing development. Using a system created at Marist College from IBM patches running on a borrowed Multiprise 3000 mini-mainframe, engineers used the SuSE AutoBuild tool to create more than 400 software packages in the first weekend. They adapted YaST (the installation, configuration and systems management tool inte-grated with SUSE Linux Enterprise) and edited package selection for mainframe distribution.
But building, installing and running Linux on the mainframe did not ensure its acceptance. Starting with System/360 in the mid-1960s, IBM had positioned its flagship computing platform for stability, reliability and backwards compatibility. To match that culture, the normally dynamic Linux code base was frozen, maintained to ensure ongoing hardware and software compatibility and positioned with subscription-based services. Customers bought into – and purchased – the concept and, with that, the first enterprise-ready and fully supported Linux operating system for the mainframe was born. It has closely tracked mainframe hardware and software progress from S/390 through various generations (z900, z990, z9, z10) to today’s premier system zEnterprise. Along the way, IBM introduced the Integrated Facility for Linux (IFL), a special-purpose System z central processor exclusively for Linux workloads. IFL runs at full System z speed, does not incur IBM software charges for traditional System z operating systems and middleware and upgrades at no cost to new technology generations.
Over the years, as underlying system architecture evolved, SUSE Linux Enterprise Server exploited complementary features, such as the File Hierarchy Standard to accommodate 64-bit adaptations, larger address spaces and coexistence of 32/64-bit applications within Linux instances. It has also been the first distro to support new machine instructions as mainframe microarchitecture advanced. Recognizing mainframe sites’ need for ready-to-run software installations, the Starter System for SUSE Linux Enterprise Server for System z – a complete pre-built installation server to streamline and simplify provisioning virtual Linux servers – has also been made available.
Linux is mature and ready for the big time
Linux is reliable, secure and efficient. In fact, its unofficial middle name could be “mission critical” consi-dering its combined heritage of enterprise-proven, industrial-strength mainframe security and thou-sands of widely used industry-standard Linux applications. A valuable third component is a robust worldwide Linux on System z community providing mutual support and skilled staffers of all levels.
This should come as no surprise. Mainframe Linux represents natural evolution of the system’s long and productive use. The beauty of that growth is that “Linux is Linux.” Staffers familiar with Linux on smaller platforms will comfortably use and support it on big iron and applications/services hosted elsewhere can generally be consolidated and scaled up to exploit mainframe reliability, availability and serviceability.
Installation and management tools galore
System z virtualization – extending four decades of development and productive use – allows ultra-flexible workload and resource partitioning by hardware (LPARs, logical partitions) and software (z/VM). These technologies allow extreme sharing of system resources, provisioning resources where and when they’re needed, over-committing real resources, handling demand spikes and virtualizing resources not present in real hardware.
Whether or not the oft-cited mainframe skills shortage exists, IBM and the Linux community have created a library of cookbooks and tutorials providing resources, plans, checklists, procedures and tools for quick and reliable mainframe Linux implementation. Once installed, Linux servers and applications can be monitored, measured, tracked and tuned to meet specific organization and application performance requirements. Native mainframe management tools and Linux-specific software agents combine to present both system-wide dashboards and fine-grain application/server details.
Why mainframe vs. x86?
Underutilized resources are uneconomical and do not pay for themselves. x86 platform CPU use rates of 10 to 15 percent are commonly tolerated as unavoidable. Additionally, virtualization factors of 10x are considered an achievement in hardware economies. Yet generations of mainframes have consistently run productively at near-100 percent CPU usage while virtualizing servers by factors in the hundreds. System z hosting configurations dramatically reduce physical server footprint, power/cooling configurations and expenses, software costs for fewer cores/processors and staffing requirements.
IBM’s newest mainframe, the IBM zEnterprise System, more closely integrates differing system architectures than has previously been possible. This allows workloads on the new IBM zEnterprise 196 mainframe server as well as workloads on select IBM POWER7 and System x blades to share resources and be managed as a single, virtualized system. The IBM zEnterprise Blade Center Extension (zBX) and the IBM Unified Resource Manger creates multiple server images and allow efficient resource allocation and utilization. SUSE Linux Enterprise from Novell runs the first workload optimized offering for zBX, the Smart Analytics Optimizer, which speeds complex analytic workloads at a lower cost per transaction.
“The new IBM zEnterprise System represents a bold move to fundamentally change how data centers are managed,” said Carrol Stafford, vice president, IBM System z. “The new mainframe is not only the fastest enterprise server in the world, it also represents a giant leap forward in how to integrate and automate workloads across heterogeneous platforms with unrivaled reliability and security. SUSE Linux Enterprise Server plays a key role across this architecture in helping clients take advantage of this performance as they aim to increase data center efficiency and consolidate workloads. In addition to exploiting Linux on System z, customers can manage and integrate workloads on selected IBM Power and System x servers through zBX powered by zManager.”
Mainframes are affordable
Smart management has moved from simple price tag comparisons (TCA, total cost of acquisition) to strategic TCO (total cost of ownership) evaluations. Long-term mainframe system value means that for many workloads, a System z provides better return on investment. While Linux gained its reputation running cheaply on commodity hardware – and, in fact, it is a great choice on any hardware – running Linux on mainframe hardware economizes on energy, floor space, software and staffing.
To simplify acquiring and configuring mainframe Linux, Novell and IBM partnered to provide the Solu-tion Edition for Enterprise Linux for System z, which is complemented by the SUSE Linux Enterprise Consolidation Suite for System z. Subscription pricing includes:
• Reduced total cost of acquisition (TCA)
• Ability to consolidate .NET workloads on IBM mainframe and .NET development plug-ins into Microsoft’s Visual Studio to develop applications on x86 and test/deploy them on System z
• Linux training, including vouchers for extensive on-demand Data Center Training Library
The truth is out there
Businesses, government agencies and non-profit organizations must consider their unique workloads and applications to determine whether System z and Linux suit their IT needs. Good application candi-dates for Linux hosting include existing apps that are well instrumented, apps using mainframe-resident data (e.g., z/OS or z/VSE), apps coordinating processes running elsewhere and apps offloading time-consuming processing to IFL cycles.
Unfortunately, “myth-information” about Linux abounds; but it’s easy to rebut uninformed, often self-serving and, of course, contradictory objections to evaluation such as…
• Linux is too new to be trusted
• Linux is obsolete, replaced by newer and better technologies
• Linux requires unique/scarce/arcane/expensive skills
• Linux is unreliable
• Linux is unsupported
• Linux and z/OS are enemies
• Linux information is scarce and too hard to find
• Linux is hard to install
• Open source software – including Linux – is evil and un-American
• Open source software is dangerous because of its chaotic/informal development environment
In any field of endeavor – woodworking, cooking, gardening, or IT – success requires the right tools. IBM’s System z is uniquely suited for hosting today’s demanding hybrid multi-system applications, ser-vices and cloud computing. It is quite clear that increasingly, Linux enriches the mainframe’s ecosystem and positions it to satisfy future requirements.
Meike Chabowski is a product marketing manager for Enterprise Linux Servers at Novell. Her responsibilities include Linux for retail and Linux on the mainframe. She can be reached at Meike.Chabowski@novell.com.