Allowed latency values between dynatrace Components

The Performance Warehouse, DT Server, and Analysis Server should all be in the same network. 

The amount of calls to the Performance Warehouse (like most databases) is large enough that best practice is to have them close together.  The Analysis server, while called infrequently, is meant to receive very large files and likewise I would recommend having close-by (e.g. not over WAN).  Does this make sense?

Rick B

Hi Rick,

Thank you for the information!

We plan to establish our data center on two different locations - the distance between the locations generates a latency of about

50 ms - therefore I need to know, if it is possible to connect the collectors over the WAN to the DT Server, when we place all

dynatrace components (Perf Warehouse, DT Server, Analysis Server) on one location.

Or is it recommended to duplicate these components for each datacenter?

Claus

Hi Kyle,

ok, that means that I can connect the collectors and DT Server over the WAN (50 ms) as long as agents/collectors and PerfWH/DT Server/Analysis Server are on the same subnet?

Is this correct?

Claus

Yes that is my understanding the only latency that really matters is the collector agent one, in the documentation there is a section on that here Collector Configuration about mid way down

"Collector Placement Close to Agents

A Collector should always be placed "close" (in terms of latency) to the connected Agents, as there are start-up round-trips between them in the thousands – to instrument application classes,... A Collector should be placed in the same sub-net / at least in the same data center as the Agents / instrumented applications, then connecting to the Dynatrace Server over the WAN, if needed."

As far as the Performance Warehouse/Server/Memory Server I'm not sure it really matters, richard.boyd@dynatrace.com would know more information about that.

Hi,

Just to add to that: between agent and collector latency is the most important factor (hence them having to be close to each other).

For Collector - Server connection the availability of sufficient bandwidth is far more important whereas latency is not. As long as the connection is fast enough to handle the amount of data that collector needs to send to the server, there should not be a problem. Data gets compressed and bulked together at the collector level to reduce the amount of round trips.

As Rick has already stated, the Performance Warehouse should be in the same network as the Dynatrace Server. The server sends many calls to the PW and does so continuously. This is not only a Dynatrace AM best practice, but is also valid for any software solution which makes frequent calls to the database.

For the Memory Analysis server this is less important though. It is an on demand service which generally is not used very frequently. Since the data is being sent over in a continuous flow, latency is not as critical. Also there will be no impact wrt monitoring data if it takes somewhat a longer time to send the memory dumps. Again bandwidth will be more important as the files are larger but the round trips are less. However, if you have a choice, but it close to the server .

Hope this helps.

KR,
Kristof 

Hi all,

thank you for your responses, the informations are realy helpful.

Regards

Claus