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
I used ray to deploy a client actor and a server actor for inference. In most cases, the response delay of the client calling the server function via remote is tens of milliseconds, but sometimes it suddenly occurs longer than one second, the amount of data requested each time is almost the same. Why is this? In addition, when I deploy services through the cluster, there are generally no timeouts between actors on the head node, but there are many timeouts on the worker nodes. Is there any additional overhead for communication between actors on worker nodes?
The text was updated successfully, but these errors were encountered:
I used ray to deploy a client actor and a server actor for inference. In most cases, the response delay of the client calling the server function via remote is tens of milliseconds, but sometimes it suddenly occurs longer than one second, the amount of data requested each time is almost the same. Why is this? In addition, when I deploy services through the cluster, there are generally no timeouts between actors on the head node, but there are many timeouts on the worker nodes. Is there any additional overhead for communication between actors on worker nodes?
The text was updated successfully, but these errors were encountered: