...
Memory usage for all nodes:
Flame graphs
The following graphs were captured on 2 of the nodes using async-profiler with the following settings:
-agentpath:/opt/async-profiler/lib/libasyncProfiler.so=start,event=cpu,loop=10m,file=profile-%t.html"
These were in the middle of a 100k caliper test set to 105 TPS for each of the 4 caliper instances (totalling 420 TPS across them). Each of the 2 nodes the flame graphs were captured for had a single caliper instance connected to them.