Automated Capacity Management Overview
ThruPut Manager provides Automated Capacity Management (ACM) facilities in addition to the automation and optimization features available through Service Level Manager. Installations using or considering sub-capacity pricing can now constrain or defer low importance batch as a peak approaches and resume normal service when it passes, generating significant monthly software savings with no impact to online service or mission critical batch.
Generate Real Savings
For intermediate-sized installations, monthly software fees can be $500,000 to $1,000,000 and up. We estimate these installations can save $50,000 to $100,000 per month using ThruPut Manager to manage their processor usage for yearly savings of up to $1 million. Larger installations can save substantially more.
Manage Processor Usage
Installations use IBM’s sub-capacity pricing when they want to treat a CPC as smaller than it is, for the purposes of software pricing. This pricing model is based on the rolling 4-hour average (R4HA), that is, the CPU usage averaged over the previous four hours. The software charges for the month are based on the highest R4HA for the whole month.
ACM continuously monitors the rolling 4-hour average for each LPAR and, optionally, LPAR group, comparing it to a “capacity target” which can be:
- the Defined Capacity for an LPAR, or
- the Group Limit for an LPAR Group, or
- a capacity target defined to ThruPut Manager when the installation does not use soft-capping.
When the rolling 4-hour average approaches the capacity target, ThruPut Manager automatically constrains or defers lower importance batch, as specified by the installation, and correspondingly relaxes these constraints as the average decreases.
Specifying Workload Constraints
For each subset of batch workload identified to ThruPut Manager, the installation specifies limits by capacity threshold for the number of executing jobs, length of time in queue before a job can be selected, or assigns a service class to the job that limits the processor consumption. ThruPut Manager automatically takes the requested actions to reduce or eliminate the contribution of low importance batch workload to the peak value of the rolling 4-hour average. Since ThruPut Manager gives preference to your important workload, you can now afford to run even more constrained, knowing that only less important work will be impacted. This means you no longer have to size your machines for the monthly peak to ensure your important workloads are not compromised.
Reduce MLC Costs With or Without Capping
The underlying philosophy of ACM is to defer less important batch workload to reduce or eliminate its contribution to the peak value of the R4HA. ACM adds more criteria to the job selection decisions being made by ThruPut Manager. With ACM configured and active, ThruPut Manager is in a unique position to control batch workload by taking into account the value of the R4HA relative to any limit defined by the datacenter, providing real savings on monthly software license fees.
ACM is designed to reduce the performance impact of “hitting the wall”, as the R4HA approaches the soft cap limit. However, not every installation prefers to put a soft cap limit in place. You can still have the best of both worlds, however, as ACM can be provided with a virtual limit that it uses for managing batch workload in relation to the R4HA. Performance remains high while the peak R4HA remains low – and your MLC costs remain correspondingly low.