In February 2020, IBM announced IBM Z Performance and Capacity Analytics V3.1 to replace IBM Z Decision Support for Capacity Planning V2.1. Read the announcement letter for more information.
On May 14, IBM announced a new pricing model for z/OS called Tailored Fit Pricing. This change represents the most radical update to mainframe software pricing for 20 years and eliminates several inhibitors for clients to understand the true cost of their workloads. The impact of switching to a Tailored Fit Pricing model affects how clients should view workload management, optimization and future capacity planning leading to opportunities for growth.
To support this transition, in particular for IT Operations Managers, Performance Analysts and Capacity Planners, new dashboards have been added to IBM Z Decision Support for Capacity Planning providing visibility into overall MSU consumption and at the job level.
The side-effects of operating under the Rolling 4-Hour Average
Until now you will have needed to become familiar with the Rolling 4-Hour Average (R4HA) pricing model to understand z/OS software costs. This can be challenging for systems administrators, capacity planners and software financial analysts as you look to understand the impact of existing and new workloads to minimize your monthly software usage bill.
Under R4HA, a key part of the optimization process meant the use of capping – effectively limiting the amount of work that could be run within the peak period to keep costs under control – and many software vendors developed offerings that assist in this regard. However, the side-effect of capping meant that trade-offs need to be made in terms of workload prioritization.
At the most basic level, capping results in increased CPU delays which in turn leads to longer response times and maybe SLA goals missed, as lower priority work is displaced by more important work. Taken to the extreme, capping could lead to operational problems because locks and other resources are not released in a timely fashion preventing the current workload to perform optimally. The key observation being that you may find the role of planners and system programmers is to focus on “How can I optimize my R4HA?” instead of optimizing the actual workload performance to minimized consumption.
Considerations for managing workloads under Tailored Fit Pricing
Transitioning to a Tailored Fit Pricing model, in particular the enterprise consumption model (where there is a baseline committed amount of MSUs), will significantly alter how you look at performance and costs. It does not eliminate the need for systems programmers and capacity planners to analyze performance, instead it empowers them to make more insightful observations that can help quantify existing workloads for chargeback and give a clearer forecast on future new workloads costs. Controlling consumption means understanding the workloads and their behavior on the system. By comparing actual system behavior with past data leads to insights about deviations in consumption.
Some questions that now need to be considered include:
- How am I tracking against my annual baseline?
- What is the change in MSU consumption month-to-month?
- Can I quantify the cost of growth workloads or impact of new business opportunities?
As everything running on the machine now counts towards the consumption usage there is value in looking at the cost of every application and workload to identify inefficiencies regardless if it previously was contributing to the peak or not. Examples may include ensuring application are using the latest compiler technologies to reduce MIPS use, or ensuring that all zIIP-eligible workloads are running on those engines instead of general processors.
Managing consumption under Tailored Fit Pricing using IBM Z Decision Support for Capacity Planning
One way of tracking consumption is using IBM Z Decision Support for Capacity Planning (IZDS for CP). IZDS for CP is performance reporting and forecasting tool. It collects and curates SMF data in near real-time providing data tables reflecting Key Performance Metrics that can be reported on to analyze performance and form the basis for future operational needs with capacity planning. Out of the box report are provided on multiple platforms, including Cognos Analytics, Splunk and the Elastic Stack.
IZDS for CP already provides support for tracking workload performance priced under the R4HA. To coincide with the launch of Tailored Fit Pricing, IZDS for CP has been enhanced to provide insight for users who have switched to enterprise consumption model (see APAR PH11764). The advantage of IZDS for CP is that the low overhead continuous curation process enables conclusions to be drawn on recent data without the need to for scheduled manual batch loading of SMF data. New dashboards enable you to understand their current level of MSU consumption (for example, tracking the consumption at a high level to see how you’re tracking to our baseline) and then drill-down quickly and efficiently to determine the workloads that are contributing to changes in the workload.
Let’s consider a few examples to illustrate how IZDS for CP addresses some of the enterprise consumption model:
In the above screen, we are showing a dashboard indicating the Total MSU % utilization of Enterprise Containers ENTERPR1 and ENTERPR2, from the start of the Tailored Fit Software pricing agreement (1 September 2018) up until the current date. It also displays the total number of lapsed days into the agreement (251).
We can click on container ENTERPR1 to automatically drill down to the more detailed level as shown below. For example, the Cumulative MSU usage for Container ENTERPR1 per month, from the start of the Tailored Fit Software Pricing agreement (1 September 2018), and up until the current date. From there we can drill down to a LPAR breakdown of the consumption.
Another key scenario is understanding what are the “normal” MSU consumption levels across all workloads and see where deviations to this consumption occurs quickly. Core to identifying this insight, clients need to compare previous consumption levels per day, on all STCs and jobs with actual consumption values today.
It’s important to compare like-for-like days to understand the deviation, for example the first working Monday of every month. Using reports similar to as shown below we can quickly identify when this happens. From there we can look at hourly breakdowns and see what applications and jobs were running at that time to contribute to the MSU consumption.
Under the new model you need to consider the consumption of all applications across the day, month, week and month instead of just what might be influencing the R4HA. To minimize consumption you might consider re-compiling application with the latest compiler technology or take a closer look at zIIP-eligible workloads. Any workload that could be run on a zIIP but instead executes on the general CP is also contributing to the MSU consumption. The dashboard below shows how IZDS for CP can indicate what zIIP-eligible workload has executed and forms the basis for making an informed decision on workload management including additional hardware purchases, if needed.
Finally, IZDS for CP provides the ability to support you to predict future consumption based on past performance and algorithmic models. This forecasting capabilities can help you understand the the current trend of consumption and whether the current entitlement matches with the workload. If additional capacity is needed due to workload growth (which of course, represents additional investment in the platform) then the amount needed can be estimated accurately and a price attached to it.
Tailored Fit Pricing changes how clients will pay for and manage workloads on z/OS. Significantly, capping is no longer an option to manage costs under a consumption model. Instead you need to understand what their normal consumption is and quickly determine if there is a deviation from this normal. Optimization strategies will be based around around ensuring workloads are tuned correctly. IZDS for CP has been enhanced to provide dashboards that can empower you as an adopters of this new model to understand current consumption levels, quickly identify the workloads that contribute to the consumption, forecasting future consuption and assist with assessing any potential costs.
To learn more about IBM Z Decision Support for Capacity Planning, refer to the Knowledge Center and Marketplace. The architecture of IZDS for CP is described in detail in an earlier post that shows the simple end-to-end collection, curation and reporting process. Please reach out to your IBM representative if you have more questions on Tailored Fit Pricing and IZDS for CP.
Thanks to Matthias Bangert and Karla Bester for their input into this article.