FHIR Chat · Question · FHIR at Scale (FAST): Exchange with/without intermediaries

Stream: FHIR at Scale (FAST): Exchange with/without intermediaries

Topic: Question

Heath Dinkins (Jan 15 2021 at 20:28):

Hey guys. I was a part of the SANER track, specifically. I saw your presentation/demo. That was interesting stuff. I had two questions:

Have you guys discussed the effects large data sets (possible long open connections) might have on this data flow?
Have you guys discussed what happens when errors occur at a hop (e.g. an intermediary node goes down) and how retries might work?

Patrick Murta (Jan 15 2021 at 20:50):

Hey Heather... thanks for the participation and questions. Here are my thoughts. 1) For large data sets, synchronous APIs are not always a good idea but the bulk data FHIR asynchronous model is. It can use our approach as well without holding connections open for too long 2) We have, but we plan to handle that in a separate IG for 'intermediary scale'. There are existing patterns that handle that well today that we can use.

Heath Dinkins (Jan 15 2021 at 20:55):

Patrick Murta said:

Hey Heather... thanks for the participation and questions. Here are my thoughts. 1) For large data sets, synchronous APIs are not always a good idea but the bulk data FHIR asynchronous model is. It can use our approach as well without holding connections open for too long 2) We have, but we plan to handle that in a separate IG for 'intermediary scale'. There are existing patterns that handle that well today that we can use.

1.) Totally agree. But, given the main protocol being used here is HTTP, I figured I'd ask about that. SFTP and other protocols can handle large files well enough but given we are trying to improve interoperability through FHIR, I was wondering how that might work.
2.) Ahhh, OK. I'm new to the "IG" stuff. What is the existing pattern within FHIR to handle retries? HTTP/REST requires the individual hops to implement their own handling of retries. So, I was just curious about how this might work.

Vassil Peytchev (Jan 15 2021 at 20:56):

Is there a link to the presentation for those who were not part of the Connectathon? Or maybe a video recording?

Patrick Murta (Jan 15 2021 at 20:57):

Yeah, the Bulk FHIR IG is meant to handle exactly what you are describing so as not to have HTTP timeout issues

Patrick Murta (Jan 15 2021 at 20:58):

Yes, @Dana Marcelonis will be able to publish that Vassil

Heath Dinkins (Jan 15 2021 at 20:58):