Scalability in DDS and Dealing with Large Systems

[@jezmerchant-locke1316]

Can you please explain a little more about the experiment you carried out? According to the DDSI-RTPS 2.2 specification a domain can have a maximum of 120 participants per domain, yet you say that base stations have 100 readers and 100 writers and satellite stations have 100 writers and 200 readers. Am I misunderstanding the use of the term domain participant from the specification?

[@zettascaletech]

Thanks for your question, Jez. Below you can find Lex's response. The ddsi spec states that there is a limit of 120 participants per node per domain. The reason for this limit has to do with the maximum amount of available UDP ports. However, this limit only applies to participants, not to endpoints (i.e., readers and writers). In DDS, a single participant can have many endpoints. For example, you can create a single participant, and within that participant, you can create hundreds or thousands of readers and writers. I think CycloneDDS uses a theoretical maximum of 2^24 endpoints per participant. In practice, you will probably never need it, and if you want to do so you will probably never reach it because it is likely that you don't have the available memory resources. I think what might be behind your question is a misinterpretation of what a participant is. In DDS, a participant conceptually represents an access point to a shared data space. Typically, participants communicate with each other by exchanging ddsi messages. Readers and writers (which are called endpoints in the specification) represent publishers and subscribers of data. Readers and writers must be created within a participant, so a participant is basically a container for a bunch of readers and writers. Because participants are aware of each other, data published by a writer on one participant ends up at a matching reader on another participant. I hope this clarifies a bit. Let me know if you have any additional questions.

[@jezmerchant-locke1316]

@@zettascaletech Thank you very much for clearing this up. This will help me going forward. I now hope that the specification would get changed to make that clearer.