Server hardware consolidation can deliver advantages beyond hardware savings. For example, application servers that run on the mainframe derive advantages from the outstanding Reliability, Availability, and Serviceability (RAS) of the mainframe hardware.
Figure 17-1 illustrates simple consolidation, where discrete servers on multiple machines are mapped to the same number of Linux images on a single mainframe. The z/VM virtualization allows you to run numerous Linux images concurrently.
Figure 17-1. Simple server consolidation to the same number of Linux images
The number of installed images remains the same, but you might be able to reduce hardware administration and maintenance costs. You can also bring the images under the control of the mainframe environment's tools.
Virtualization leads to further advantages. Machines that run single applications often have a low average utilization. Applications that run in z/VM guests consume significant resources only when there is work for them to do. Figure 17-2 shows the consolidated servers of Figure 17-1 at a single point in time when the load is different for each application. In the diagrams, the sizes of the boxes indicate the relative resource consumption at a particular time.
Figure 17-2. Linux images can scale with the work load
In effect, z/VM dynamically scales the Linux images' CPU resources according to their load. This scaling is based on z/VM's virtualization. Adding CPU in a distributed environment involves hardware changes. Scaling the virtual CPU resources in a Linux-on-the-mainframe environment is transparent to the application and does not even require rebooting Linux. The z/VM guest definition determines the range within which this dynamic scaling is possible and permissible.
If you want to make more resources available to an application, you can scale it vertically by changing the respective z/VM guest definition. When you do this, you enable z/VM to assign more resources to it. You can also scale it horizontally by deploying additional Linux images to run more instances of the application. Deployment of new Linux images is very fast, because under z/VM images are logical definitions and can share the already available hardware resources.
What if your mainframe machine runs short of processing power for all the scaling that is going on? There is yet another level of scaling. You can also scale the capacity of a mainframe machine. Visit http://www.ibm.com/servers/eserver/vertcap.html for information on the Capacity Upgrade on Demand options.
Server hardware consolidation is likely to be profitable in environments such as that of ISPCompany where there is a large number of servers to be consolidated.
StoreCompany also takes advantage of server hardware consolidation by moving its departmental servers onto the mainframe. For example, it plans to consolidate all its Samba print and file servers on the mainframe. Figure 17-3 illustrates the consolidation scenario.
Figure 17-3. Consolidating Samba servers
StoreCompany not only gains flexibility from Linux on the mainframe through consolidation. It also uses Linux to quickly bring additional applications to its mainframe environment, while assuring close integration with applications on the traditional mainframe operating systems. This translates into improved time-to-market for new projects. We look at integrated server environments with Linux applications and applications on traditional mainframe operating systems in Chapter 19, "Building Integrated Server Environments."
Simple server hardware consolidation is an opportunity for implementing a successful Linux-on-the-mainframe project and to realize some savings. It is also an opportunity to gain experience and confidence with Linux on the mainframe and to pave the way for future more complex and profitable projects.
The final answer on whether a hardware consolidation makes sense for a particular environment can be given only after a TCO analysis. It typically examines various options for providing the hardware. For low-utilized infrastructure servers (for example, DNS servers or archive file servers), there might be enough spare capacity on an existing machine. Where significant workload is moved to Linux on the mainframe, you might need extra processing capacity. Often it is advantageous to purchase dedicated IFL features for the migrated workload. See 21.4, "Integrated Facility for Linux (IFL)."