@abradley201 this is ready for review/testing/merging.
NOTE The throughput for sql and sync is shows to be lower. My fix on the benchmark suite was more focused on correctness of the benchmark logic. So The numbers we get here are probably almost what the previous author intended to reflect on the benchmark, but couldn't because os misuse of go concurrency primitives.
The number 10 of goroutines was found through trial and error. The main focus was to run the benchmark in the same machine. Voltdb works in memory, add replication to three node it is easy to run out of memory. So with 10 go routines I was able to successful run sql,sync and async in reliable manner.
The memory leak for async is gone, the culpit was mishandling of draining of connections.
@abradley201 this is ready for review/testing/merging.
NOTE The throughput for sql and sync is shows to be lower. My fix on the benchmark suite was more focused on correctness of the benchmark logic. So The numbers we get here are probably almost what the previous author intended to reflect on the benchmark, but couldn't because os misuse of go concurrency primitives.
The number
10of goroutines was found through trial and error. The main focus was to run the benchmark in the same machine. Voltdb works in memory, add replication to three node it is easy to run out of memory. So with 10 go routines I was able to successful runsql,syncandasyncin reliable manner.The memory leak for
asyncis gone, the culpit was mishandling ofdrainingof connections.