How well is your mainframe outsourcer managing capacity and performance?
Part 4: The outsourcer’s cost structure
This is the fourth in a series of blogs about how to follow up on your mainframe outsourcer. Previously I’ve discussed why it is important to follow up on your outsourcer, understanding MIPS and MSU, and the pitfalls of MIPS. In this article we will start looking at the commercial aspects related to paying for CPU capacity. Specifically, we will look at some of the components that are rolled into CPU prices and then try to shed some light on the outsourcer’s own cost drivers. In the next blog, we will dig into CPU based pricing models.
What do you pay for?
Mainframe outsourcing agreements cover much more than just CPU power. They typically also cover other forms of capacity such as storage (disk and tape), memory and network. Software may be included in the CPU costs or may be billed separately. Service Level Agreements, business continuity requirements, security requirements, complexity (e.g. number of LPARS, CICS regions, etc.), location (e.g. in country or offshore), reporting requirements, and many other special terms and conditions also affect the price paid. ‘Soft’ requirements like agility and design authority can have a large impact on the price and the success of an outsourcing engagement but are difficult to write into a contract.
Many of these cost elements are typically rolled into the CPU price, so I will focus on CPU pricing considerations in this blog. It is common for the CPU price, normally based on MIPS or MSU, to include some software and services and reflect some special terms such as location, SLA, and business continuity requirements. This makes it difficult to do a direct comparison of MIPS or MSU prices between outsourcers.
As discussed earlier, having a clear definition of what measurements and calculations form the basis for the billing is essential. It is also critical to ensure good reporting and access to performance and capacity data (SMF) from the outsourcer in order to be able to do an independent validation of the charges.
What does your outsourcer pay for?
Outsourcers have a lot of different costs, some of which are passed on directly and some of which are rolled into the CPU cost. These costs include hardware, software, staff, housing, cooling, electricity, and other infrastructure. The most important variable cost for the mainframe outsourcer is IBM software such as zOS, DB2, CICS, IMS. IBM bills for software based on CPU utilization. For many years the most common model was to pay for IBM software based on the monthly peak rolling four-hour average MSU utilization. See my previous blog for an explanation of this measurement. More recently IBM has introduced ‘Tailor Fit Pricing’, where software is billed based on total MSU usage during the month rather than the peak. Tailor Fit Pricing is not generally available to outsourcers at this time, but we have seen exceptions and this rule may change over time.
IBM prices their software on a curve. The first MSU’s are very expensive, but the unit price falls rapidly as the number of MSU’s increases. This is why outsourcing makes sense. By gathering many smaller customers on one mainframe the outsourcer gets a significant volume discount and can sell CPU power at a much lower price than the customer would pay on a stand-alone machine.
Understanding what the outsourcer is paying for can be useful when negotiating with the outsourcer. If they are paying for the monthly peak usage, and that peak is during the morning of the first Monday of each month, then anything running outside of that period is basically running for free from and IBM software perspective, though obviously there are other costs to cover. We encourage an alignment of interests and an open dialog between the outsourcer and the customer on this topic. If both parties can work together to move workload away from the monthly peak, then there is a basis for a win-win.
CP, zIIP, IFL
It is important to understand the various processor types when discussing CPU based billing. Originally IBM only had one processor type – general processors. These are referred to as ‘GP’, ‘GCP’ or more often just as ‘CP’. They can run all workloads and IBM’s software billing, typically the most expensive part of the mainframe, is based on CP usage.
Two other types of processors, Integrated Information Processors (zIIPs) and Integrated Facility for Linux (IFL), are physically the same as a CP, but the functionality has been limited through microcode changes. zIIP’s can run certain types of DB2 workload, but also other stuff. IFL’s can run Linux, but not zOS. zIIP’s and IFL’s are a commercial rather than a technical invention – the outsourcer doesn’t pay for IBM software running on zIIP’s or IFL’s. They were created by IBM as a way of providing a discount on certain newer types of workloads while continuing to protect IBM’s revenue on legacy workloads.
Alignment of Interest
We encourage customers to look for a win-win with their outsourcers. Understanding what is included in your CPU costs and how they relate to the outsourcer’s own costs is an important first step. In the next article, we will dig into pricing models in more detail.
by Steven Thomas, Chief Technology Officer, SMT Data