Mainframe Outsourcing Transparency

Outsourced mainframe customers often lack a clear understanding of how well their outsourcer is managing the mainframe capacity and performance relative to what is optimal for the customer. The customer seldom has access to the data needed to really understand their capacity usage and often lack the skills to communicate clearly with the outsourcer on capacity and performance questions. The outsourcer may have limited motivation to help the customer optimize capacity costs. In many cases, the customer doesn’t even have the ability to validate whether the outsourcer is invoicing in a correct manner relative to the contract or in a fair manner relative to industry ‘best practice’. Simple questions like ‘how is the basis for invoicing (e.g. MIPS) calculated?’ can lead to a 10% difference in the invoice.

Many companies outsource their mainframe operations. And billing models vary – from customers who pay for a fixed capacity based on capping, to customers who have more complex models such as paying for the 90th percentile peak MIPS usage between 8:00 am and 5:00 pm weekdays. Often the actual contracts are negotiated by people on both sides with limited insight into the technical nuances. Once the contract is signed, other people take over the actual measurement and billing, and we often see a disconnect between what was agreed and what is invoiced. This could be because the agreement leaves room for interpretation – for example how MIPS are calculated. It could also be because the outsourcer’s configuration changes in ways that are not foreseen in the contract. E.g. if the outsourcer adds zIIP engines to the mainframe, this will add an overhead that increases the MIPS utilization of all users. Is that effect regulated in the contract?

Some of the important questions that a customer and outsourcer should have a common understanding of include:

• Does the capacity usage (for example in CPU seconds) on the invoice match the measured usage as seen in the SMF Data?
• Has the outsourcer used the agreed methods for calculating the capacity for which the customer is paying (for example MIPS calculated from CPU seconds)?
• Are the agreed methods for invoicing capacity in line with industry ‘best practice’? For example, is the MP overhead of additional processors fairly reflected in the MIPS rate used to bill the customer?
• What are the advantages and disadvantages of other methods of determining capacity usage for the customer and the outsourcer?
• How can the billing model contribute to an alignment of interests between the outsourcer and the customer? For example, if the outsourcer’s costs are driven by peak monthly MSU usage, then how can both parties be motivated to move workload away from that peak?
• What are the cost drivers, where is there optimization potential and what specific actions can the customer or outsourcer take to achieve savings?
• Are there capacity related performance issues such as poor response time, and what options are there for addressing these besides increasing capacity?
• What are the ongoing reporting requirements to ensure a transparency between the outsourcer and the customer?

Answering these questions requires access to data and knowledge of the technical aspects of the agreements between the customer and the outsourcer.

The most relevant data is typically SMF 70, which gives the resource usage at the LPAR level, and is what invoicing is typically based on. Other SMF data such as SMF 72, 30, 101, 110 will allow a deeper understanding of the actual cost drivers seen from workloads, jobs, Db2 and CICS respectively. This kind of data is essential for understanding how to reduce cost or optimize performance.

The prices or commercial terms from the contracts are not actually necessary to start a transparency discussion. What is interesting is the technical definitions. E.g. Is the number of CPU seconds on the SMF 70 record used as a basis for the billing? How is the usage normalized across different machine configurations? E.g. how are MIPS calculated from CPU seconds? How are peaks determined? Is there a difference in the way CP and zIIP MIPS are billed?

Opening these discussions can be a delicate exercise between the customer and the outsourcer. To avoid contention, it is important to be fact based. What is measured? How are the calculations done? What is billed? What can be done to reduce the bill? Clear and transparent answers to these questions are the basis for a positive working relationship between outsourcer and customer.