Systems management, like all other aspects of a business, is subject to economic constraints. A significant contribution to the TCO of an IT infrastructure is the number of people required to manage it. The number of FTE (Full Time Equivalent) employees required not only depends on the size of the installation, but also on the diversity of the hardware and software and on the desired characteristics, such as security and availability.
For example, ISPCompany runs hundreds of Linux images and expects more images as its business grows. It is inconceivable to manage all of them without the use of tools. Generally, the more numerous and uniform the systems to be managed, the more beneficial it is to use tools.
Let's make the absurd assumption that a large number of systems are managed individually and that all interventions take place through a command line at the respective system's console. Assuming that one FTE can handle a given number of systems, the cost of maintaining the systems is roughly proportional to the number of systems to be maintained. This would mean that administrators would have to be hired and trained at the rate at which the systems increase. Consequently, the potential for growth would be capped by the number of available administrators.
Tools increase the efficiency of administrators and thus keep the number of FTEs down, but are themselves associated with cost. The cost of a new tool is not only the purchase or development cost but includes more hidden aspects like deployment, training, and maintenance. Typically, the expenditure for a tool is high at the time of introduction and comparatively low while it is in use. Applying the same tool to an increased number of systems might mean higher license fees, but it usually means a less than proportional increase in overall costs.
Figure 12-1 illustrates the cost characteristics of people-dominated system administration versus tool-dominated administration. Introducing zSeries hardware and Linux as new elements into an IT environment is likely to mean that either existing tools must be adapted or alternative tools must be acquired to handle the new environment. Tools that accommodate Linux on the mainframe are already available. Linux has been designed as a simple system to manage; that is, it does not come with too many control points that you have to take care of. On the other hand, you can apply system or kernel patches to include any additional controls you need.
Figure 12-1. Cost of system administration for a growing number of systems
Looking for new tools implies the chance to build in the capability for handling a growing number of systems. For example, our hypothetical StoreCompany initially introduces only a few experimental Linux images. As StoreCompany gains confidence and experience with Linux, implementing the departmental servers across the company on Linux could be an attractive option. If the tooling put in place for the first experimental Linux images is chosen with that in mind, moving 50 or 100 departmental servers to Linux images can be an easy exercise. Moving to Linux on the mainframe is also an opportunity to examine whether the tools you are using can be replaced by something that is more suitable for what you want to attain.
Because systems management must be driven by business needs and not by the capabilities of a tool or set of tools, it is essential to know exactly how the business needs translate to management tasks. We argue that this is best achieved by first mapping the business objectives to policies and then to procedures. With procedures at hand, you can then assess how suitable a particular tool is for your purpose. Before we do that, we will map out the territory of systems management.