Comparing Obelisk with Temporal and Restate
4 points by tomas
4 points by tomas
I'll admit I didn't have much idea what this did until I had a certain thing explain it to me. Basically it sounds like a fault tolerant system for apps and it lets them continue when you have a failure.
The only thing we know about our code is there will be random runtime failures. What if you could restart from just before the failure, instead of the very beginning?
Wondering why the sleep for one second is included. It seems so far two separate examples sleep and then just wait for the answers anyway.
It is just meant to show the gist of workflow APIs: starting child executions, adding durable delays, and awaiting results. I agree that the sleep does not make much sense in this particular workflow.
I think the go to marketing example is to deactivate a user after a 30 day trial period by just durably waiting for 30 days.
Cool, thanks for explaining.
Near project, do you see Obelisk as competing with the other serverless WASM-running platforms?
I don't really see it as a competition, it is OSS after all. I started with a vague idea about what needed to be fixed, what tool I would have liked to exist when recalling many of the challenges I experienced at my previous jobs. Those requirements shaped the project: ability to see exactly what happened, not just logs but the whole execution with stack traces, centralizing concepts like retries, cancellation, deployment, failure recovery. The split between pure logic and effects allows all of that.
I don't mean business competition, more like mindshare. When I'm looking for random places to put serverless functionality, do you think this fits the bill of "serverless runner that happens to be durable execution"?