i don't think there is a limit on number of concurrent requestors one node can handle. Requestors contains thread and clipboard information, which consumes your heap. I will make sure that you configure enough heap for your JVM to handle the number of requestors you anticipate to avoid OOM.
This depends on your the sizing of your app servers(JVM/Memory) and the average clipboard size of requestors in your application. There is no general guideline for this and this sizing should be determined primarily based on the number of peak requestors your application has to support and their average clipboard size.
Using this the number of app servers/JVMs, Memory, CPU size will be calculated.
Thanks Waikei and Sethu. How can we estimate the average clipboard size of a requestor. I am developing a Service-REST where at certain point of time I Pega can receive around 300 to 400 requests , which will lead to creation of around 200 to 300 requestors
The service activity will create a work object and send a response with the case id.
Access the SMA Application during your peak volume window and click the estimate size button. This is available in SMA-->Requestor Management, on the RHS Section at the top left( under all requestors).. Clicking this will add a column called Datasize in the list view with the size of memory occupied by your requestors.. You shall take an average of this.
It is recommended to do this in a lower environment/stress environment first, then gather metrics around this and then forecast the hardware requirements(Memory,CPU etc) based on your proposed transaction/end user volume.