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
In the current state sync implementation, every node in the network will attempt to obtain state parts in roughly the same order (modulo any variation due to parallelism of buffer_unordered):
The peer selection mechanism is designed such that requests for the same state parts will be routed to the same hosts. The motivation is that we can benefit from caching of parts on the host side.
We should shuffle the order of requests on the client side to avoid generating spikes of requests to the same hosts.
The text was updated successfully, but these errors were encountered:
In the current state sync implementation, every node in the network will attempt to obtain state parts in roughly the same order (modulo any variation due to parallelism of
buffer_unordered
):nearcore/chain/client/src/sync/state/shard.rs
Lines 85 to 98 in 4e87bae
The peer selection mechanism is designed such that requests for the same state parts will be routed to the same hosts. The motivation is that we can benefit from caching of parts on the host side.
We should shuffle the order of requests on the client side to avoid generating spikes of requests to the same hosts.
The text was updated successfully, but these errors were encountered: