Hello capacity planners and networking friends!

This is part two of a five-part series focused on network capacity planning using IBM’s SevOne network observability solution. 

Over the coming weeks we will cover the following topics:

• Planning for capacity with TopN views (May 9)
• Using projection trending (Today’s Topic!)
• The value of histograms when planning for capacity (May 30)
• How are percentiles calculated, and what value do they provide? (June 13)
• Generating capacity reports using RIA Workflows (June 27)

Let's get into today's topic -- Projection Trending – What is it?  Why is it important?  How is it used?  Where is it found in SevOne?

What is it? 

When planning for network capacity, one of the tools in our toolbox is Projection Trending.  Projection trending gives us the ability to project a trend line into the future to allow us to estimate where we might end up 30/60/90+ days from now based on the growth (or lack of growth) that has occurred in the past. 

Why is it important?

When planning for capacity, it is important to use the data at our disposal to reasonably approximate how much capacity we will need in the near future, so that we don’t get caught with a capacity shortage.  If you remember from last week’s blog, there was a quote that likened network bandwidth to oxygen for your applications.  We don’t want to run out network bandwidth (our oxygen) or we will starve our applications of the expected performance, causing unnecessary delays and unsatisfied customers.  Planning for capacity takes time – analyzing links, the traffic driving the bandwidth utilization, getting capacity recommendations approved, coordinating and provisioning the changes with NetOps and the carriers involved, and validating the changes after the fact emphasizes that we want to get things right the first time so that we don’t suffer unwarranted delays or premature or unexpected shortages.

How is it used?

When planning for capacity, the capacity planner must consider:

1.        Which data and how much history should be included?

a.        Has traffic grown relatively consistently over time? 

b.       How does seasonality come into play?

c.        Are there changes that have happened recently that need to be factored into the amount of history we decide to reflect on?

2.        Are there time windows where traffic is much lighter (non-business hours) vs. peak business hours?

3.        How far into the future are they trying to estimate future bandwidth requirements for?

Based on these considerations, capacity planners can do “what-if” analysis to see how the different settings impact the results, allowing them to estimate what they need.

Where is it found in SevOne?

TopN Widget

Within the TopN Widget, you can specify a number of settings to take advantage of projection trending to estimate future needs based on past performance.

Forward Looking Timespans – You can specify one or more future timespan settings (next month, next 3-months, next quarter, next 6 months, etc.) to see where your capacity will be at.

Device Work Hours – This is important.  Don’t water down your projection by including off-peak data.

Days to Threshold – When you select a forward looking timespan setting and choose the option to sort based on that value, you can use the “days to threshold” option to see how many days until you reach your threshold value.

The TopN result tables highlight the top interfaces that are projected to be > 80% utilized in the next 90 days:

Performance Metrics Widget

Within the Performance Metrics Widget, there are several settings that allow you to use SevOne’s historical data to project a trend line into the future.

When you enable the trend option in the widget settings, you can then set the following:

Trend type – “Projected Trend”

Days to project – Enter the number of days to project into the future (e.g. 90)

Lookback time – You can determine how far back you want to look (how much historical data to consider) when you calculate your forward-looking linear regression projection line.

Algorithm – Linear is the most commonly used algorithm.  This generates a “line of best fit” by using the historical data to perform linear regression analysis on the data.

Work hours – You will want to select “Device” or your own custom work hours windows here so that you consider utilization during times of significance vs. off-hours, so you don’t water down your projection data.

Here is a 90-Day projected trend example showing the resulting 90-day forward looking trend line for an interface:

Summary

Today we reviewed Projection Trending—what it is, its importance, how it’s used, and where to find it in IBM’s SevOne network observability solution.  I hope you were able to learn something new that you can apply to your own capacity planning efforts!