Many customers are asking us what the difference is between Group Capping and BMC Intelligent Capping. In this blog, we'll make this clear.
The key difference between BMC Intelligent Capping and group capacity capping is their behavior when the MSULIMIT/Group Limit is reached. If the limit is not being reached, you are not saving money, so what occurs at the limit is extremely important.
With Group Capping:
- Capping decisions are based on LPAR weights, not workload importance. This means high-importance work can be delayed, while low-importance work still runs.
- LPARs make the decision to cap themselves based on their weight, and there is no consideration of what and how much is running on other LPARS. This leads to an inefficient use of capacity.
- When capping occurs, no further LPAR weight changes are made until capping ends. This can take hours and result in high-importance work being delayed.
With BMC Intelligent Capping:
- Capping decisions are made based on workload importance to protect critical work by ensuring that capacity is available in the right place at the right time.
- LPARs in a policy are truly managed as a group, and decisions are made with full visibility to all LPARs, allowing it to make intelligent decisions about allocation of MSUs.
- When capping occurs, Intelligent Capping continues to dynamically adjust the MSU allocations to best meet the needs of the LPARs and the high-importance work.
BMC Intelligent Capping is designed from the ground up to save monthly license charge (MLC) costs and reduce business risk.
The importance of importance
With group capping, as long as the LPARs are collectively below the Group Capacity Limit (GCL), it works reasonably well. However, when you begin to approach the group limit (GCL), things don’t work as expected–and that’s a big issue. Maybe you don’t usually approach the limit? In this case, there’s no real purpose in using a group, because it is not saving money. In a well-configured system, the limit should be reached often– that’s how you save money. This is where Intelligent Capping excels over group capping.
In a capacity group, when the GCL is reached, one or more of the LPARs will be capped. Which LPARs will depend on the weights of the LPARs in the group. LPARs in a group are allocated a share of the group’s limit based on their weight relative to other LPARs in the group. Capacity groups don’t care what is running on an LPAR (they are not importance aware). If it is using more than its share, an LPAR will be capped – even if it’s running all high-importance work! Meanwhile, you might have an LPAR running an unimportant batch job that continues to run. This is not what you want to have happen.
Another problem with capacity groups: each LPAR in the group independently decides if it needs to cap itself. LPARs are not aware of other LPARs in a group and whether they are using all of their share or what they are running. The decisions made are pretty limited and are not very efficient.
Some customers use IBM Intelligent Resource Director (IRD) to automatically adjust LPAR weights in response to WLM SLA goals to try and make Group Capacity more efficient, but this doesn’t help as much as you might think. IRD only manages weight adjustments between LPARs in a single Sysplex on a single CEC. This is referred to as a Cluster. So, if there are LPARs in the group that are not in the Sysplex, the effectiveness of IRD is reduced. Even worse, as soon as an LPAR in the group is capped, IRD will leave the cap alone or reset it to the initial value. In either case, no further weight changes will occur while it is capped, which might span many hours. During that time, drastic changes in workload Importance can occur and will go un-managed.
The Road to Intelligent Capping
With Intelligent Capping, the LPARs in a policy are truly managed as a group and decisions are made with full visibility to all LPARs, allowing intelligent decisions about MSU allocations. Intelligent Capping monitors each LPAR to determine how much work is running, as well as the 4-hour rolling average, the importance of the workloads, and the overall MSU target for the policy. Decisions are made based on workload importance, and LPARs with higher importance workloads will receive MSUs from those running lower importance work. This protects critical work by making sure that capacity is available in the right place at the right time, regardless of whether the LPARs are all in the same Sysplex or not.
As with Group Capacity, a well-configured system should reach the Intelligent Capping MSULIMIT, saving you the most money. By using Intelligent Capping to cap LPARs, it continues to actively manage the LPARs and their workloads and continues to dynamically adjust the MSU allocations to best meet the needs of those LPARs and the high-importance work. This allows a lower MSU limit to be set, which reduces MLC costs. In contrast, lowering MSUs by reducing a group’s GCL means that the same percentage reduction will be applied evenly across all LPARs in the GCL. There is no awareness of workload importance and it assumes that every LPAR can afford to donate an equal percentage of their MSUs. This will likely result in even more high-importance work being capped, and therefore not an acceptable cost savings option.
Intelligent Capping policies also enable you to provide advanced criteria to further maximize savings and reduce risk. This includes the ability to consider the cost of MSUs – a patent pending technology from BMC Software.
In summary, Intelligent Capping is designed from the ground up to reduce your MLC costs and reduce risk. This is not the case with WLM capacity groups. For a test drive, download the MLC Savings Estimator and enter your data to see how much Intelligent Capping can save you. Click here to Download the BMC MLC quick utility
Content contributed by Paul Spicer, Lead Product Manager, BMC