We are considering minimizing effort for first mvp deliver of a a solution implementation where an external system will invoke a web api to start a process which will run a series of unattended nodes activity chained through to return the response in a synchronous manor.
Its my understanding reading community posts and Appian documentation that this is feasible but is it a bad idea/practice to do so in the Appian architecture? Has anyone every done this with success? If so what is count of requests per month you are handling and how long does your process take to complete? What other aspects should I consider before going down this path? Our solution will be deployed in Appian Cloud environment.As noted the main reason to consider this is mvp delivery planning but if its a bad idea then want to plan appropriately. The variables I can not assess now is how long the process will take to execute since we have not built it yet. For load we need to handle 30K requests per month initially and business obviously wants a solution which will grow way beyond current needs.
Discussion posts and replies are publicly visible
Not having detailed insight into your process: Your plan sounds fine to me. 7.5k transactions a week could be fine.
Note - whatever is going on in your process ought to be fast, and iron clad (i.e. very unlikely to break). If it's slow, and error prone, it's probably not a good fit for a synchronous response. Your alternative (and it sounds like you thought about it already) could be to kick of an integration call from Appian to your external system after the processing in Appian is done, and send a "response" payload that's logically linked to the initial request.
Hi Robert, thanks for your insights, totally agree with your concerns with regards to process design requirements; we had similar concerns in internal discussion before I created this post. Until we get rules fully understood and developed just do not have enough details now to know complexity or how long a single execution will take. If they are too costly then we will have to do asynch model.
Do you have any insights to offer with regards to how well Appian cloud environment would scale with sync vs. async models? I am assuming asynch would be better but by how much?
Hey Bob -
It's possible that there's a difference in how these two designs will impact scalability. I can't really tell you off hand what that would be.
Your line of questioning brings to mind a couple of playbook articles that I think you'll want to review:
Hi Robert thank you.
© 2021 Appian. All rights reserved.