Elementary Capacity Management - Trending and Forecasting

Thursday, May 05 2011 07:42 Written by VMGuru

This second of my four part Capacity Management series talks about the concept of understanding Trending and Forecasting as it relates to Capacity Management. The first, and most important concept that we must understand is the concept of utilization. Understanding utilization is vital to the success of any cloud or virtualization project that is undertaken. Without a clear grasp on the current utilization of an environment, there is no way to trend historical data for the purpose of accurate forecasting of the future.

For those that missed Part 1 of this series which served as the introduction, I'm going to strip technology out of this piece and provide more simplistic examples of what capacity really is. This will allow these concepts to be applied to any technology stack, not just virtualization.

Utilization

Utilization is the metric that measures how much of the environment is being used at a particular point in time. This is the simplest form of capacity reporting, as it normally deals with only a single point in time… Right Now. This provides a quick snapshot by stating, “If I look at my environment right now, how much capacity do I have, and how much of it is utilized?”

Once we have an understanding of how much more is available, we can start to create advanced use cases within the constraints of the current point in time.

How many total objects will my container hold?
How many objects am I currently holding?
How many more objects can I add?
If I want to add X more objects, will they fit?

The image below represents a wooden crate that can hold a specific number of apples. While not all apples are grown to the same size, after years of picking apples I know exactly how large the average sized apple is and how that relates to the crates that I use to carry them.

Utilization

Referencing the diagram above, we can quickly answer the listed capacity questions.

My crate can hold a total of 6 apples
My crate is currently holding 4 apples
I can add 2 more apples to my crate
If I want to add more than 2 apples, I will not be able to and either need a second crate

By calculating the value of Capacity / Usage, I can calculate my utilization of my crate to be (4/6) or 66.6% utilized. This means I have (2/3 OR 100% - 66.6%) 33.3% of my capacity available

Utilization is important to understand what is happening at a finite point in time. It is absolutely required to understand the basics of Capacity Management, but by itself, can provide only so much depth into capacity. In order to really understand what is happening, we need to take a closer look at trending.

Trending

The concept of trending introduces the dimension of time into the world of capacity. Trending is required to determine the rate of chance within an environment. This gives people the visibility to understand how their utilization has changed over a period time. Calculating the trending is required before we can start doing any forecasting of how we anticipate utilization to change moving into the future. We use trending analysis to answer one simple question:

How is my utilization changing over time?

For this example, let's assume that my apple business is now booming and I've brought on a new employee. With all new employees, there is a ramp up time and learning curve involved in the job. We'll call our new employee...Steve.

Trending

Let's assume that during his first week, Steve was able to bring 2 apples back to the shed. Definitely not great, but again, there's a pretty steep learning curve to find the perfect apple for my high-margin business of selling perfect apples to rich people with too much money. During his second week of work, Steve managed to bring 3 apples back to the shed, an improvement over the previous week. During his third week, which happens to be the current week, Steve managed to bring back 4 apples to the shed.

We need to take a look at multiple snapshots of utilization over a specified period of time. By looking at these changes over time, we are increasing our utilization by one unit ever week. Since this is a linear pattern, we can use some basic calculations to understand the slope of the line (m = ΔY/ΔX). With such a simple example, all this is going to tell us is that our slope is 1, meaning we are adding 1 additional apple to the crate each week.

It is important to note that within a more realistic example such as virtualization, there will rarely be a linear growth trend due to external factors (adjustments), so more advanced calculations may be required to improve accuracy of the trending analysis.

Now that we understand what our trend is, we can wrap up today's post in talking about Forecasting

Forecasting

Once we have calculated the growth trend of a specific property, we can start to analyze this data for the purpose of forecasting what will occur in the future. Forecasting methods will always be estimates of what may occur in the future. There are different methods that can be used to gain better accuracy, but it is simply not possible to state that we can forecast exactly what is going to happen. This is due to the fact that there is likely always adjustment occurring in the environment and each adjustment modifies linear path trending.

Forecasting helps us answer an additional 2 questions as it relates to capacity:

How any objects will I be consuming at a point in the future?
When will I run out of capacity?

It is critical to understand that forecasting is ALWAYS an estimate. There are different ways to calculate for greater accuracy, but there is no way any forecasting model will ever provide 100% accuracy.

For the purpose of the definition, we want to take a look at the following diagram.

Forecasting

Being very happy with Steve's productivity to date, I decide to keep him on board. I can see the weekly improvements in his work, and can start to make some estimates about his productivity increases in the future. Knowing from the trending analysis that he is adding 1 additional apple to his crate each week, I can anticipate that next week he will bring back 5 apples, and in 2 weeks, he will bring back a full crate of 6 apples. There are several factors that can cause a shift in the trend, and ultimately adjust our forecast, and we will call those instances Adjustments, and will talk about them in the next post in the series. I am able to forecast this information using a linear equation based on the slope that was calculated in the previous example. (y = mx+b). Through some basic calculations I can choose a point on the trend line to figure out the value of b. Once i have b, I can predict what any Y value (number of apples) will be based on the current week (x). Thinking back on this ALMOST makes me regret being one of those kids in school who boldly stated "When am I ever going to use this crap in the real world?"

Utilization Forecasting

Trending and forecasting make up a bulk of the information required for Capacity Management. Unfortunately, not everything exists in a Utopian world, and Adjustments and Tuning are required. Make sure you come back next week when we dig deeper into what happens when things don't remain consistent!