Checkout the new results which make ScyllaDB shine even more than before.
The test was performed on the same hardware as the previous benchmark.
The main change was the increase in the number of loader VMs. We’ve noticed that this change shifted the throughput saturation point, allowing to squeeze even more tps from ScyllaDB by loading the system more.
The table below summarizes the change from the previous results. The score is average CQL operations per second (rounded up):
Workload | Old result | New result | Improvement |
---|---|---|---|
write | 1.3M | 1.87M | 44% |
read | 1.18M | 1.59M | 34% |
mixed | 1.18M | 1.37M | 16% |