yep can be queue, database, … idea here was to show to treat machines like CPU unit, not state unit
After job is done, you can kill them all 
1 Like
good description…
they could let us to send only a beam file, with the function, creating a web view of the hot-upgrade process, so latency would comeback to soft-real time, and they would save a lot of resources.
1 Like
But I think the problem is general languages on VM.
Now java has https://www.graalvm.org/. So if you want to create lambda function using java you run first code from Gral VM and you have fast startup time. Cloud is moving from long running processes to fast kill fast start.
3 Likes
This was a helpful introductory comment into serverless and specifically lambda. I’m highly unfamiliar with the architecture, and reading through this was very informative. Thank you!
Does someone has a working lambda runtime? The official repo seems to be broken
AWS has just announced SnapStart which starts your app after deployment, takes a snapshot of the disk + memory at that time, and then spawns very quickly an already initialized application.
This looks like a good solution to the BEAM slow start time shortcomings aforementioned in this topic.
Do you think we can leverage this with Elixir?
6 Likes
