Monday, 27 July 2015

Forecasting, when modeling is not the only choice - Forecasting Techniques (3 of 5)

In my opinion there are four main techniques that are used in forecasting: analytical modeling, simulation modeling, trending and headroom charts.

Let’s take a look at each of these:
Analytical Modeling                                        

        The ability to predict environment changes based upon the arrival of work
·      CPU
·      IO
        Based upon
·      Baseline timeframe (seasonal peaks and troughs, highest transaction times etc)
·      Calibration of model – Does it match real life?
·      What-If changes to model
        Spend virtual $’s, helping you to determine what is the best cost benefit that you can get out of it
        Low cost modeling technique, as you’re able to take the data and apply it going forward without making costly mistakes.
The graph below illustrates how this can be considered, always taking in to account how the ‘end user’ will be affected.
Simulation Modeling
There are many organizations such as Wall Street companies, who feel the need to run simulation models but it does take longer to set up and can be very costly.
        Replication of environment
        Has to stay close to production environment
        Usage of tools to replicate workload into simulated environment
        Longer lead time to setup
        Provides granular detail on environment changes
        Can be costly
With trending you are usually looking at one metric for each chart, this doesn’t prohibit you from having several charts looking at different metrics for the same scenario.
        Provide prediction based upon an interval of time
        Quick to produce
        Usually looking at one metric
        Can show steps in growth changes
        Confidence is based on length of interval and amount of historical data
The more data that you have to feed in to your trend chart then the more confidence you can have in the end result.
On Wednesday I'll be taking a more in-depth look at Trend types.
Charles Johnson
Principal Consultant

No comments:

Post a Comment