As indicated in 4.3, "StoreCompany and its business application," StoreCompany is looking at setting up a new Web catalog with third-party items that StoreCompany would not have to stock. Now it is up to you as "decision maker" to decide whether to give this new idea a try. Naturally, you ask, "What is the cost of this opportunity?" Linux on the mainframe brings some interesting factors to the TCO calculation.
Total cost of ownership attempts to give a picture of total cost of an IT system over a period of time, typically three to five years. A TCO analysis is especially of interest when you are considering any significant change, such as evaluating a new business opportunity, upgrading your servers, or moving locations. This can be contrasted with total cost of acquisition, which takes into account only the cost of buying new machinery and associated software but does not include any other costs.
A TCO study is made up of a multitude of factors. Some are quantifiable, such as the typical factors listed in Table 6-1. Other factors are of a more hidden nature and are listed in Table 6-2. Some factors have a particular importance when you are considering the combination of Linux and the mainframe.
In short, calculating TCO is a complicated business, and every organization ultimately must come up with its own method.
A typical TCO equation thus sums the cost of hardware, software, occupancy, people, and services. In addition to this, you may want to consider costs of a more hidden nature such as those listed in Table 6-2.
Let us go quickly through the typical TCO factors before we look at what Linux and the mainframe bring to the equation. Note that we will not give you exact numbers for any of the parameters, because every one of these varies throughout the world. For example, labor costs are not uniform across the globe.
Hardware usually includes:
A physical server costs the same whether its utilization is 100% or 10%. Chances are that not only the test and supporting servers are low-utilized. Typically, the utilization of the production servers is in the 10 20% range. It may be worth looking into consolidating that whole set of servers on z/VM. Assuming you do not have work that requires new, dedicated capacity, the virtualization technology offered by z/VM can save you real hardware, such as CPUs, disks, and cabling.
Software cost includes acquisition, plus service and support of the programs used on the system:
Additionally, software costs generally include the backup software and essential network services such as directory services, messaging, routing security, and network management.
While it is possible to build a Linux system from sources on the Internet for free, this may not be suitable for production systems. With Linux, the operating system cost may be low, especially if the price is based on the number of processors you have (rather than the number of images). The cost of a distribution today may include maintenance.
In fact, for all of the software, you have the option of building it yourself (the cost will be in the development), buying it (and getting the service that goes with a contract), or using Open Source (some cost will be in the support, either in a support contract or as personnel cost).
6.1.3 Maintenance and support
The purpose of a support contract is for an enterprise to have access to a problem reporting mechanism (usually a phone number), including an assurance that problems reported will be fixed. Some software providers offer a single number to call for all of the products in the solution. In other words, they accept responsibility for the solution. Others cover only the products they sell.
The cost of support will be a trade-off between what you want to do yourself versus what the service contract should cover. Part of your TCO analysis might be to decide on the trade-off. It is the time of your staff versus an expense. The range of possibilities in the Linux space is wide. If, for example, you are happy with using mailing lists to find solutions to problems in Open Source programs, the cost of service can be low. You can decide on an umbrella contract that covers everything, such as IBM Global Services offers. Most Linux distributors offer support as well.
Services involve a contract or contracts with a consultant or others who are hired to do something, such as develop a tool. An example of services for Linux on the mainframe could be getting z/VM operational, if you don't have the skills available. Or, if you don't have Linux skill, you could contract that out until your people get their education. Or you could get one of the distributors to come in and set up Linux for you.
You have a lot of control over services. The question is, for the particular solution you are considering, how many services are you going to buy to get the solution up and running?
Servers must be housed somewhere. The cost of the raised floor in square feet, space for access and the control center, and power are usually included in an occupancy parameter for TCO.
Part of what used to make the mainframe a "dinosaur" was the cooling, the raised floor, and the cabling. Today, a mainframe uses less of almost everything than a typical server farm. A mainframe no longer needs a raised floor and has a moderate footprint, as shown in Table 6-3.
Given the availability of high-speed fiber for interconnecting sites, the cost of occupancy may be a less critical factor in the future. You now have more choice as to where you want to locate the equipment. However, consolidating football fields of servers on a mainframe can mean substantial savings in square feet and power (power to run the machine, to cool the machine, power for your staff working there, etc.).
If you think about a typical low-end server (such as a PC) consuming, for example, a kilowatt, and the biggest z900 mainframe consuming around 15 KW and being capable of functionally replacing numerous low-utilized servers, the cost savings in power can be dramatic.
Example: While a typical configuration of a server farm can cost hundreds of dollars a day in electricity to run, a single z900 running the same workload can be an order of magnitude cheaper. The savings are even more dramatic when the floor space requirements of a server farm are considered. The average server farm can require up to some 10,000 square feet of floor space compared with the much smaller footprint of the mainframe including its peripheral devices, as shown in Table 6-3.
Personnel cost is normally measured in units of full-time equivalents (FTEs).
Personnel typically included might be operators, system administrators, and data administrators. The personnel cost can be calculated by multiplying the number of FTEs with a current average salary.
How many machines or servers or how much data one person can handle often depends on the type of task the servers are doing. FTE averages that are sometimes used are shown in Table 6-4.
In general, the more similar the servers in a server environment are, the more servers one FTE can manage.