Virtualization and consolidation have continued to move IT workloads to the zEnterprise System and Linux has been integral to that, but as sites move forward, they’re also looking at best practices that further improve the process. Corporate IT isn’t alone in the continuous improvement effort. Linux and mainframe vendors are also providing new tools that improve virtual Linux system creation, deployment, performance monitoring, and optimization. These efforts couldn’t come at a better time, since IBM recently noted that one-third of zEnterprise sites now run Linux on z/VM.
Current challenges include:
- Speed of deployment
- Linux workload performance and optimization
- Continued eradication of IT staff “silos” in heterogeneous workload management (including Linux)
- ROI and Total Cost of Ownership (TCO)
- Virtual server sprawl.
The Business Case for Virtual Linux
If it’s to be an able player in the new private cloud IT models organizations are defining for themselves, Linux must be easy and fast to create and deploy. Private cloud IT infrastructure will be characterized by rapid on-demand provisioning of resources. Initially, this will occur manually, but as system management tools deliver new capabilities, more resource and workload provisioning will be automated. There will be provisioning for user departments; we’re already seeing some IT departments introduce chargeback mechanisms. Consider, too, the application development lifecycle. IT application programmers increasingly will demand setups of resources needed to test applications before deployment. In the Linux (and other) worlds, this will mean the rapid creation and deployment of testing and staging environments for applications on demand. The use of virtual systems is a critical element in accelerating system creation and deployment.
Linux on System z Image Creation and Deployment
More than a decade of creating and deploying virtual Linux systems in a System z environment has yielded a body of knowledge and best practices. Still, companies face challenges as the ante is raised for even faster, more efficient deployments and workload management. The challenges begin with creating virtual Linux images and extend to deployment and de-allocation of Linux systems.
For most sites, creating a virtual Linux system image in z/VM begins with copying a script originally “hand developed” for other Linux virtual system creations. Staff has a library of scripts they can pull from, and the job is to customize a script for the environment and resources on z/VM the new application will require. This process on z/VM lacks a robust Graphical User Interface (GUI), which is why technical staff use a method of copying, pasting, and tailoring scripts in image creation. There’s some point-and-click administration capability for Linux on System z, but there’s also a z/VM wrapper surrounding Linux on System z that includes the Open Systems Adapter (OSA), hypersockets, etc. Implementing Linux on System z requires z/VM mainframe skills—an important point for IT data center management, where effort has been spent the past few years breaking down “silos” between different IT skills groups (e.g., mainframe vs. distributed systems staff). The heterogeneous private cloud data centers of the future will require staff cooperation that’s as seamless as the IT infrastructure itself.
The good news is that toolset producers on both the mainframe and distributed sides of the data center are coming together with new solutions that help bridge the gap and break down the silos between mainframe and distributed personnel and skillsets in Linux. They’re doing it with new GUI-based tools that let IT staff create a virtual Linux image (even on an x86 machine) and, with the click of a mouse, target the image for either an x86 or a zEnterprise environment. Automation is built in so the tool goes out to determine the resources needed (e.g., memory, subsystems, etc.) to carry the workload for which the system image is being created. After the tool self-discovers the resources needed to run the workload, it pulls them into a menu a technician can peruse to choose what he wants. The tool will automatically identify errors whenever changes are introduced in system image creation that threaten compatibility with underlying operating system licensing and support. Early results from one vendor’s internal pilot testing of more than a one-year duration produced a speed improvement in virtual Linux image creation, testing, and deployment of seven to eight times over what time-to-deploy would have been using manual script creation, testing, and image deployment.
Workload Optimization Challenges
Effective virtual Linux system image creation and deployment transcend simply allocating the correct resources and ensuring the underlying operating systems remain compatible. Correct sizing of virtual Linux operating systems is a key issue that starts in the image creation and deployment stage and ultimately affects the optimization of virtual Linux systems and workloads.