How can you be sure the estimated max server capability is not actually just a limitation in the client, i.e, the client maxes out at sending 224k requests / second.
I see that this is clearly not the case here, but in general how can one be sure?
You parallelize the load from multiple clients (running on separate hardware). There are some open source projects that facilitate this sort of workload (and the subsequent aggregation of results/stats.)
I see that this is clearly not the case here, but in general how can one be sure?