Currently, the system is able to send SET requests with random keys to the server with the following command:
fkvs-benchmark -t set -c 10 -r -n 1000000
but the key generation happens in the hot path which is I/O heavy, generating the keys as we do I/O is not performant and affects benchmarking results, to make results more precise and less skewed we should ideally pre-generate keys during a warmup window before running the actual benchmark.
This will lead to:
- More precise benchmarking results
- Better I/O performance when performing SET with random keys
Currently, the system is able to send SET requests with random keys to the server with the following command:
fkvs-benchmark -t set -c 10 -r -n 1000000but the key generation happens in the hot path which is I/O heavy, generating the keys as we do I/O is not performant and affects benchmarking results, to make results more precise and less skewed we should ideally pre-generate keys during a warmup window before running the actual benchmark.
This will lead to: