Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Starting the Process
The modeling process begins with creative people. They real key to modeling is the people -- the package, though important, is secondary. No matter what package is used, those responsible for it need to be able to abstract the hardware and/or software being modeled and use the modeling tools to represent this abstraction.
This is the hardest part of the modeling process. You must decide exactly what is important and what's not as important from a modeling perspective for the hardware and/or software that you plan to model. Very often, the more technical you are, the harder it is to model, as everything seems too important to leave out, but for most modeling projects this is simply not an option, as you cannot model the entire hardware and software process end-to-end.
The other important decision that you must make is determining the expected level of accuracy. This level of accuracy is often a tradeoff between time, money, and management expectations. You need to sit down with management and answer:
- What is an acceptable level of accuracy?
- What is the budget?
- Can a model be created with this level of accuracy and the proposed budget?
Most importantly, everyone needs to agree on what happens if you expect an 80% level of accuracy and management buys hardware and software assuming 100% accuracy. This leaves a 20% gap, which means someone will be pointing fingers.
What I have often seen in this process is that if you can model to 90% on most large systems, then the additional 10% can be made up in just buying more hardware. Of course, there are systems that do not meet this criteria. I have worked systems that had to be modeled to 99% accuracy, as they were designed around hardware that was running at near-peak operational rate. Therefore, accurate modeling at the 99% level was required to meet the concerns the customer had at running a near-peak hardware rate.
I have to admit that I am often too caught up in the inner workings of the hardware to be a good modeler for the 80% level. I always want to add every piece of hardware and software to the model with the justification that it could be important in case XYZ. I often get lost in the details and forget the big picture. What I have found is that I am not a good modeler unless you want very accurate models. That is why I say it is a rare person that can model accurately at 80% model reliability. That person has to know exactly what is important and how to abstract the important, somewhat important, occasionally important, and unimportant parts of the hardware and software.
Once the hardware and software is modeled at the desired level of accuracy, it is on to the next step in the modeling process.