Showing results for 
Show  only  | Search instead for 
Did you mean: 

Collector Sizing and Capacity


Hi all,

This doubt is about the capacity of the collectors in Appmon 6.5.

In this document we can see the maximum agents supported by each collector size. I want to design collectors groups for my Dynatrace environment. Apart of the groups based on data centers, I want to separate the agents groups in collectors depend of MPS that generate each agent group…we have agent group of architecture process, application process, etc….
How can I see the maximum MPS supported by each collector size? What is the MPS medium used to define the maximum agents supported by each collector size? Or maybe Dynatrace used another metric to establish the maximum agent supported?

Thanks in advance.



Hello Alberto,

I would recommend you to use the Deployment Sizing calculator, follow these steps:

Once you have gone through with the above the you can ask more to get some practical examples.



Thanks Babar for your reply. I have used the Deployment Sizing
Calculator to obtain the number of collector based on my agent number. My
question is in other way, I want to group agents in collectors and groups of
collectors using the MPS as criteria to group it. The problem for me is that I cannot
locate any information about capacity of the collector I only found the
capacity in transactions of the server. It’s important for me to ensure that
the collector support the intense traffic.

Hello Alberto,

As we know that the AppMon Collector performs the byte-code modifications when instrumented applications start up and the maximum number of Collectors (60) for a single Server.

The Dynatrace Server has to handle cyclic measures generated by Agents, so you should not worry about the MPS for the collector.

Rest all the information is available in the below 'Collector sizing' link.



Ok Babar, then suppose the follow example:

We have 400 java agents in my system. Inside the recommendation Dynatrace talk about the large collector that can contains 400 java agents.
Thinking in the high availability we will install two medium collectors and we want to group the different agents groups of my system profile in these collectors. Then we are thinking in do this grouping using as a metric the MPS generated by each agent group. We use MPS because we don’t know other metric to be in account that be related with the activity supported from the agent to the collector.

Then suppose this scenario:

Agent Group 1Very high intensity of MPS (1500 MPS)
Agent Group 2High intensity of MPS (300 MPS)
Agent Group 3Normal intensity of MPS (50 MPS)
Agent Group 4Low instensity of MPS (2 MPS)

I will group the agents in two collectors:

Collector A -> Agent Group 1, Agent Group 4

Collector B -> Agent Group 2, Agent Group 3

The objective of this distribution is to have not saturation the collector A. If we use a large collector maybe we have problems with the performance because the agent group 1 can saturate the collector with the intensity of metrics that send to the collector.

For this reason we need to know other metric apart of the number of agents to be ensured that the charge of the collector is correct and discard performance problems.

Thanks a lot Babar.

Hello Alberto,

I would like to add one more thing that you should also consider the TPS along with the MPS which is most important for the correct sizing of the platform.