You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
While throughput/overhead for a "raw" test is the most important metric and this is currently covered.
GC/Allocations are also important. Especially when sockets/TLS are used in the scope of a larger application and in resource constrained environments (Docker for instance).
What we see is on servers that are under stress with full workloads, where the user code is already churning the memory GC can take a significant percentage of the processor time.
To this end I would recommend that during a benchmark there is an option to spit out GC information. This could be done at the start and end to get a "snapshot" or at each of the intervals.
If you are interested in this I could have a crack at somestage to prototype something.
The text was updated successfully, but these errors were encountered:
While throughput/overhead for a "raw" test is the most important metric and this is currently covered.
GC/Allocations are also important. Especially when sockets/TLS are used in the scope of a larger application and in resource constrained environments (Docker for instance).
What we see is on servers that are under stress with full workloads, where the user code is already churning the memory GC can take a significant percentage of the processor time.
To this end I would recommend that during a benchmark there is an option to spit out GC information. This could be done at the start and end to get a "snapshot" or at each of the intervals.
If you are interested in this I could have a crack at somestage to prototype something.
The text was updated successfully, but these errors were encountered: