h3. Architecture1:
TBD Elastic Search works in a way that it uses 2 different memory: heap and off heap.
I gave each 256 MB for those and things were fine when processing 50K sdk init events per minute. However, the memory usage reported by {{docker stats}} was 750 MB, so maybe there is some other parameter to configure.
h3. Architecture 4:
Memory usages:
* Clean bootstrap: ~100 MB
* While processing 1K/minute sdk init events: no change, ~100 MB (that's like having 15 devices starting up an application every second)
* While processing 50K/minute sdk init events: ~700 MB (that's like having 833 devices starting up an application every second)
* Bootstrap with millions of sdk init events processed: ~200 MB (probably Prometheus or something is doing some indexing/processing upon bootstrap)
I didn't really configure any memory constraints for any of the applications here.
They're all Go apps that allocate memory when needed w/o limit. I think at least Postgres has an option to limit memory usage, but I didn't configure it since other apps didn't have that. |
|
