LeaderTM is a system modeling kernel specifically designed to allow the performance evaluation of virtualized systems currently being used by thousands of companies. It boasts a modern design which allows it to capture intricacies of today’s computer systems both in the area of CPU partitioning and the area of I/O subsystems. Advanced solution methods implemented in LeaderTM pay special attention to scalability issues. The modeling algorithms used in LeaderTM have been developed by leading experts in the area of capacity planning and performance modeling with years of experience in the IT industry and research organizations in leading universities.
The general view adopted in LeaderTM is that of a set of CPU complexes arbitrarily sharing a farm of I/O devices. Each of the CPU complexes may comprise multiple physical processors divided into Logical Partitions consisting of some number of logical CPUs. The physical CPU share of each Logical Partition may be capped. LeaderTM is capable of representing specific partitioning algorithms such as implemented in z/OS. The I/O subsystem can also be represented at its several levels of abstraction, from simple average I/O time to the details of modern I/O controllers. Memory contention is represented using a proven model of multiprogramming. Workloads are represented through their corresponding utilizations of systems resources.
PALLAS uses the LeaderTM software under license for its professional services business in the area of mainframe system performance.