Re:CockroachDB beta-20161013
by aphyr on Thursday February 16, 2017 @01:35PM
(#13662329)
The clock skew limit controls the maximum latency on reads--so yeah, if you're doing reads that contend with writes, improving clock resolution improves latency. Not entiiirely sure about the throughput impact though.
[ Parent
| on HN ]
At a greater risk of losing a server in your cluster due to automatic shutdown after too much clock-drift, right?
You wouldn't reduce the clock drift limit unless you had more accurate clocks however. If all of your servers are on bare metal and synced to a high resolution atomic or GPS clock then you can probably reduce it from 250ms to 10ms or less similar to Googles Spanner setup.