Operating Systems

“There’s no more room in the data center. What are we going to do? These servers just keep appearing.”

How many of you remember saying that or hearing someone else say that five or 10 years ago?

This server proliferation occurred at a difficult economic time when building a new data center or even expanding an existing one was cost-prohibitive. In many instances, z/VM was brought into those data centers to save the day. By consolidating hundreds of physical servers onto a single or small number of System z footprints, companies could realize significant savings with regard to floor space, cooling, power and administration expenses. Business growth could continue without expanding the data center floor space.

Jumping forward to 2010, we were faced with another new challenge—virtual server sprawl. The ease of configuring and provisioning virtual servers on z/VM allowed system staff to implement new Linux servers in the blink of an eye; unlike physical servers, virtual servers require less space, fewer cables and less time to configure. However, now, instead of the physical data center being at maximum capacity, we started stretching the bounds of the virtual data center, the z/VM system itself.

For more than a decade, one of IBM’s vision statements for z/VM has been: “Making systems practical and profitable for customers through virtualization and its exploitation.” Such a vision is what started IBM down a multiple release approach to expand the limits of z/VM, beginning with z/VM Version 6 Release 1. While not a large release in terms of lines of code, it laid the foundation by introducing a new architecture level set of the System z10 processor family, allowing z/VM Version 6R1 and future releases to exploit new instructions and facilities with minimal use of dual paths (see Figure 1). This made the system more maintainable and extensible.

The next step with regard to virtual server relief came in z/VM Version 6 Release 2 with its introduction of Single System Image (SSI) Clusters and Live Guest Relocation (LGR). This support was made available through the priced VMSSI feature of z/VM. The SSI Clustering capability provided horizontal scaling by connecting up to four z/VM systems into a cluster where the general virtual servers could run on any of the four systems. Management capabilities were enhanced to decrease the effort of managing these four systems, compared to managing four individual systems.

For example, you would install service just once for the cluster instead of individually on each system. Ordinarily, there would be concerns in growing systems larger in terms of dealing with planned outages for software or hardware changes. Enter LGR, which allowed a virtual server to be moved between two z/VM system images in the SSI cluster while it was running. This relocation, which is transparent to the application, helped deal with those planned outages by giving you a way to move all work off a given z/VM system without stopping the virtual servers. SSI clustering and LGR increased scalability and the possibilities for solving business problems.

z/VM Version 6 Release 3: A Return to Vertical Growth

While the horizontal scaling of z/VM V6R2 was a departure from increasing the limits of a single z/VM system, IBM hasn’t forgotten how to do that. The next step in scaling is found in z/VM Version 6 Release 3 with a return to vertical growth. This newest release became generally available on July 26, 2013. There are three core areas in the release that, when combined, focus on delivering value to the customer by:

• Reducing the number of z/VM systems required to host a large number of virtual servers
• Improving memory management flexibility and processor efficiency
• Simplifying z/VM systems management.

2 Pages