Development of a capacity planning methodology
Original Publication Date: 2005-Nov-11
Included in the Prior Art Database: 2005-Nov-11
A meta-method is disclosed, that can be used to develop capacity planning methods. To illustrate the usage of this meta-method, a capacity planning method for a business process management system (BPMS) like IBM WebSphere Process Choreographer (WPC) is developed.
Development of a capacity planning methodology Development of a capacity planning methodologyDevelopment of a capacity planning methodology Development of a capacity planning methodology
It is proposed that even for complex software products no fine grain measurements for the elementary steps of complex tasks are performed, but only rather coarse grain (minimal impact) measurements of the time intervals from the start of a business task to it's end. Then with adding knowledge about the way the involved software product(s) perform the elementary steps one can form a model, which allows to determine the desired cost factors for the elementary steps from the coarse grain measurement results by applying mathematical analysis methods such as least squares optimization (sometimes also known as minimization of variances).
This meta-method consists of 4 major steps.
The first step is to develop a cost model based on (a) knowledge about the way a system works internally (e.g. it's principles of operation) and (b) a subject matter oriented decomposition of the input to the system . This cost model typically contains several unknown cost factors, that describe the processing cost (e.g. execution time) of the building blocks that are the result of the above decomposition .
Application to a Business Process Managing System (BPMS)
A business process is something that is instantiated from a process model , consists of several steps, which the BPMS's navigation engine is navigating through. Such processes can be simple or complex and any kind of capacity planning has to take this complexity into account. A natural way of estimating of the total processing cost b of a single process instance is to use a linear combination of the cost factors x(i) that are associated with the principles the BPMS uses to run a single process instance from the beginning to the end.
These principles translate to the parameters a(i) that are multiplied with the cost factors x(i). And the sum of all these products for a single process instance represents the total cost b for this process instance .
Let's assume the following processing characteristics for the BPMS it can handle 3 different types of activities denoted as t1, t2, t3 there is a fixed lifetime cost for a single process instance , denoted as x1, independent of the amount of activities the cost for a single activity of type t1 is denoted as x2 the cost for a single activity of type t2 is denoted as x3 the cost for a single activity of type t3 is denoted as x4 there can be parallel execution paths in the business process and the cost for each additional execution path is denoted as x5
Considering a process model (named model_1) with 5 sequential activities of type t1. Then a cost model for the way this is processed in the BPMS engine could look like: - one time fixed process instance cost (independent on the number of activities), denoted as x1 plus - five times the navigation an...