15. Capacity Forecasting

Last Updated OnMarch 31, 2022

Introduction

Using Capacity forecaster you can

View a trend analysis of transactions on an application, including month-on-month growth trend at a cumulative and transaction-wise level. You can view corresponding trends on breach metrics (in terms of anomaly count) to see workload mixes or clusters that give rise to maximum anomalies.
Generate trend forecast report on a particular duration of data calculating growth trend for transactions in the specified duration.
Enter transaction growth forecasts on a particular transaction mix and generate a fine-grained capacity forecast report on all applicable metrics.

Benefits of Capacity Forecasting

You can observe capacity trends corresponding to workload signatures. You can arrive at the exact workload mix causing maximum resource contention. For example, a combination of high CPU Util, high memory utilization, and high I/Os. It helps you plan for capacity based on projected growth of this particular mix.
Capacity forecast report helps in reduction in number of problems occurring due to capacity breach and also in costs of running data center by optimizing currently over-provisioned services.
You can generate capacity forecast report to plan infrastructure and configurations ahead of time to prevent signals.

Application Summary

An Application consists of Services that communicate with each other. Each Service is a logical representation of a specific type of Component like Web, App, or DB. A Service is also a “cluster”, a logical grouping of multiple homogeneous instances.

Capacity Heat Map for the Application

Click View Capacity Heat Map to view the Capacity Heat Map for the application. It is a weekly trend which displays the frequency of Infrastructure or Application level capacity coming in proximity of the configured resource limits. Only Capacity KPIs are used for making Capacity baselines. Each Capacity KPI represents some resource limit in the infrastructure or code of the application.

Capacity KPIs are grouped under following categories –

Infrastructure Level Capacity
- CPU – KPIs related to CPU consumption
- Memory – KPIs related to Memory consumption
- I/O – KPIs capturing I/O levels
Application Level Capacity
- Application – KPIs capturing virtual resources like thread pool, DB Connection Pool, DB locks, and DB Sessions.

	Name of the application for which Capacity Heat Map is displayed.
	Heat Map is generated with respect to the timestamp when CF baseline was last generated and completed successfully. Default is duration of months for which CF baseline was generated last. You can select the months to view data – 1, 2, or 3 months. Say, CF baseline was performed for 2 months duration, application loads data for 2 months. In this case, baseline is from the date and time displayed to last 60 days. While performing CF baseline for 1, 2 or 3 months, if data for any month is not available, then application displays 0 data for that week.
	A mix of transactions of various types is called Workload Mix. Workload Mix displays volume growth over the data points as per the selected time range. This displays composition of top 10 transactions in stacked bar graph.
	Size of the bubble : Capacity KPIs have maximum and minimum limits. Size indicates the frequency i.e. the number of times a KPI value came close to the maximum during a week. When you hover on the bubble, HEAL displays five buckets of frequencies. First is the high severe bucket. Last is the least severe bucket.
	Color of the bubble : Color indicates the intensity i.e. closeness of the KPI value to the limit. It is represented as five shades of Red. Normal value is represented as Green.

Hover on the information icon to view the baseline run months as displayed in following screen.

Capacity Heat Map with Limited Data

Hover on the Total Volume graph displays the date and average volume for that day across all the transactions in a week.

You can select first, second, or third month with respect to the date and time as displayed in following screen.

If MLE baselining is not done for a specific month, that month is disabled as displayed in following screen.

Click on a bubble in the heat map to display a pop up as seen in the following screen. This pop up displays –

List of Services for the application. You can select a service to view its details.
List of all the instances that are part of this service.
A breakdown of heat for each host and component level capacity KPI that are factored for this capacity limit.

Hover on a bubble at the instance level displays a pop up as seen in the following screen. This pop up displays –

Name of the service instance
Host or Component level KPI name
Frequency of the KPI

You can select a particular transaction mix for a deep-dive analysis on that particular week.

This displays top 10 transaction names in stacked bar graph for the selected week. Hover on a transaction displays average volume for the week for that specific transaction. It also displays the average response time in milliseconds for the transaction.

Hover on a partial transaction name in the Workload Mix displays full name of the transaction as follows.

Hover on the information icon in Mix section displays following screen.

Hover on the information icon in Capacity Categories section displays following screen.

Generating Capacity Forecast

Click Generate Capacity Forecast for this Workload Mix to generate Capacity Forecast for a particular week. This displays following screen.

You can select a workload mix as follows –

Select From a Week : This generates a list of top “taxing” transactions from the selected week.
Select For a Capacity Metric : This generates a list of top transactions that cause most “stress” on the selected Capacity KPI.

Based on the selection of Workload Mix, corresponding list of transactions are loaded. Spark line displays system calculated growth trend on each transaction. Projected Peak Throughput is based on current volume of transactions and percentage increase in transaction volume.

You can select a Capacity KPI of interest as shown in the following screen.

You can enter growth plan for any of the listed transactions. Alternatively, you can enter an overall growth plan for all transactions. You can specify a percentage increase in transaction volume. You can tweak the growth percentage using the slider.

Click Generate Capacity Forecast for this Growth Plan. Application loads a Capacity Forecast heat map at the bottom of the screen. You can edit the growth plan and click Generate again for latest Growth Plan. This refreshes the Capacity Forecast heat map.

Hover on a KPI in the heat map displays a pop up as seen in the following screen. This pop up displays –

Name of the instance
Name of the metric
Current value of the metric
Projected value for the metric

MLE Automatically Predicting Transactions Trends

Allow MLE to identify transactions Automatically slider is default disabled. If you enable this slider, then MLE automatically predicts the trend for transactions. You need not drag the slider in Projected Growth column to input the transaction volume. MLE automatically predicts the peak throughput for the transactions. MLE predicts the projected growth for the transactions as below.

Projected Growth % = ((Projected Peak Throughput – Current Peak Throughput)*100)/Current Peak Throughput

Based on the duration you select, MLE automatically predicts the percentage growth for the transaction based on the historical trends.