You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The default of 3 coordinators becomes a bottleneck in large high tps databases that acquire a lot of read-only snapshots. Too many nodes direct traffic to a single coordinator that has to send replies to other nodes, and IC pool of a single node becomes overloaded, with event delivery time going to hundreds of milliseconds and even seconds.
Reconfiguring coordinators for a running database is difficult to do, and may require multiple releases to ensure backward compatibility (so unlikely to happen before 2026), but it may be useful for scaling large databases in the future.
The text was updated successfully, but these errors were encountered:
The default of 3 coordinators becomes a bottleneck in large high tps databases that acquire a lot of read-only snapshots. Too many nodes direct traffic to a single coordinator that has to send replies to other nodes, and IC pool of a single node becomes overloaded, with event delivery time going to hundreds of milliseconds and even seconds.
Reconfiguring coordinators for a running database is difficult to do, and may require multiple releases to ensure backward compatibility (so unlikely to happen before 2026), but it may be useful for scaling large databases in the future.
The text was updated successfully, but these errors were encountered: