Monday, October 20, 2008

Managing Complexity (Capacity 1)

_

There is perhaps no problem I encounter more frequently than that of managing complexity. When working with systems that operate on thousands if not millions of executions, there always exists complexity, both implicit and explicit, and for a system to succeed (meet user requirements) this complexity must be managed.

I was recently asked to determine the capacity of throughput for a certain system. To do so it was necessary to:
  1. Create a list of low level list of functions\assumptions\constraints\requirements (depending on who you're talking with the term will vary).
  2. Create a high level list of functions\assumptions\constraints\ requirements
  3. Create a series of tests to functions\assumptions\constraints\ requirements
  4. Execute tests to determine validity of functions\assumptions\constraints\ requirements
  5. Repeat as required
I thought this project was an excellent example of managing complexity. The process that was used in this case to answer a question about a system of complexity is quite typical, and the pattern can be applied to any system. The idea is to start with an entire system, identify manageable high level components of the system, and break down each high level component to a series of low level components. This cycle of breaking down can take place as many times as necessary until what was once a complicated system, is just finite number of simplicities.

ETS Capacity

For our purposes, capacity can be defined as the function of maximizing output (test data) from input (test procedure). The question then, in this case, is “how effective is our system at generating output?”. The answer can be found by compartmentalizing our system into manageable pieces and identifying how each piece behaves, both individually and with other pieces of the system, to effect the generation of test data. The diagram below displays the high level component breakdown of ETS:



The diagram reveals that the test procedure acts as an input to the system, which executes for the purpose of generating test data. Secondly, the diagram reveals that starting at the point of data generation (PXI command execution), each component is dependent on all of the components contained therein. For example, recipe definition is dependent on PXI command execution, scheduling is dependent on recipe definition and PXI command execution, and each component is dependent on the pieces contained within it. Finally, the diagram reveals that the entire system (input, components, and outputs) is time dependent. That is, time is the single most effective measure of how efficiently the system as a whole (and each of its components) is running. Thus, any investigation into the capacity or efficiency of the system should begin with defining the expected performance of each component with respect to time. Only after assumptions for each component are clearly defined can the assumptions for the system as a whole be realized.

No comments: