Mainframe computers are used in the daily operations of large organizations to perform critical and complex applications. Since they play a crucial role in the business operations, you need to ensure that you get best mainframe computer for your business needs.
But how do you determine the capacity of mainframe computers, which is directly proportional to the software licensing costs and is essential in disaster recovery planning or business continuity planning? Here are some popular methods used to determine the capacity of mainframe computers.
The Millions of Instructions Per Second (MIPS) is an approximation used to measure the computing performance of large servers such as mainframe computers.
However, MIPS calculation does not consider speed of I/O devices, speed of information transfer between memory and CPU, processor architecture, and the like, which influence the mainframe performance. Thus, MIPS is not considered as an exact measure of mainframes’ performance. Further, MIPS may vary considerably as the CPU performance varies from one processor to the other.
Mainframe manufacturers will not mention MIPS rating in the Large Systems Performance Reference (LSPR); they mention Processor Capacity Index (PCI), which is almost similar to MIPS.
MIPS is still used by many manufacturers and vendors as rough guide to determine the computing cost. The value of the mainframe is calculated in MIPS per dollar. Thus, for a typical mainframe, the more the MIPS delivered, the better the value for the money.
The conventional way to determine the performance of a mainframe computer is to record the number of CPU seconds consumed by each unit of work. However, as the work done in one CPU second is not same for each processor, this method of calculation is considered less accurate.
To address this issue, Service Units (SU), which are calculated by multiplying CPU seconds of a mainframe by a constant value, were developed. This value is called as SRM (Systems Resource Manager) constant. Mainframe manufacturers publish SRM constant for each processor in LSPR.
With the help of SUs and the rate of CPU usage, MSUs (Millions of Service Units per hour) are calculated which help in determining the capacity of the processor. MSU is a reasonable approximation of system capacity. MSU rating is mostly used by vendors to determine software licensing fee.
Internal Transaction Rate (ITR) is a popular method used to determine the capability of mainframe computers with respect to CPU speed. ITR can be simply defined as number of transactions made by the machine per CPU second (number of seconds CPU is in use).
Mainframe manufacturers publish ITR in the product’s LSPR. We can compare this number to the ITR value of the base processor, which has a pre-defined value, to determine the capacity of the mainframe. ITR is one of the reliable methods to measure the capacity and performance of mainframe computers.
Though all the three values are found in the mainframe manufacturer’s LSPR, the assessment will be made in unconstrained environments for specific workloads and thus the measurement varies for each processor. However, for a rough approximation, you can use these methods based on your requirement like – MIPS for easy calculation, MSU to determine software licensing costs and ITR for capacity planning. For accurate calculations, use tools provided by the mainframe manufacturers.