I have a cluster of five nodes.
The Replica factor value is set to 3.
Then, load test was performed by setting the insert and select ratio equally.
At this time, the response time was 8 ms for insertion and 4 ms for selection. (When 2000 TPS)
Next, the Replica factor value was set to 5.
At this time, the response time was 3.5ms for insertion and 2.7ms for selection. (When 2000 TPS)
The Replica factor was larger, but it was performed faster. Why?
I checked the Admin UI Dashboard to find the reason.
I found something strange on the hardware dashboard.
On Replica 3, Disk Wirte Ops was lower than when Replica 5. I think it is normal thing.
However, Read Ops was rather higher than when Replica 5. This is strange thing.
If there are more nodes than the Replica factor, do more disk reads?
In common sense, in the case of clusters composed of the same number of nodes, if the replica factor grows, the write latency should also be increased. But that’s not. Why?