caliper
caliper copied to clipboard
Warm-up period for benchmarks without waiting time
Currently, there is always a sleeping period (of 5 seconds) in between multiple rounds of the same benchmark. When conducting benchmarks, the first couple of transactions are likely to yield better performance when all nodes are initially idle, than when the network has reached its full utilization.
A useful feature could be to allow one to initialize the network's utilization, for instance by having a warm-up round, but without the idle/sleeping time in between those rounds. This would allow one to configure a round with, say, X transactions to warm up the nodes, only after which the measurements will be taken.