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 Core 6.x and 7.0 dose_main has a function CanSend, that attempts to make it so that pool distribution does not starve "normal" traffic. This function actually has the reverse result, in that it makes it very easy to starve the pool distribution if there is a fair amount of normal traffic while the pool distribution is going.
It is also quite easy to starve out different types of traffic in dose_main by for example having one app create lots of entity updates, and another send requests slowly. In this case the slow requests may never get through at all.
dose_main (or communication) needs a mechanism to treat all types of data reasonably fairly. Some kind of round-robin between different data type identifiers and also pd and normal traffic (which I think currently use the same data type identifier).
Pd and normal traffic may need to have different data type identifiers, if the data type identifiers is the basis of how we solve this issue.
It is also not completely clear whether this mechanism belongs in communication or in dose_main. We want to avoid adding latency to the data transmission.
The text was updated successfully, but these errors were encountered:
In Core 6.x and 7.0 dose_main has a function CanSend, that attempts to make it so that pool distribution does not starve "normal" traffic. This function actually has the reverse result, in that it makes it very easy to starve the pool distribution if there is a fair amount of normal traffic while the pool distribution is going.
It is also quite easy to starve out different types of traffic in dose_main by for example having one app create lots of entity updates, and another send requests slowly. In this case the slow requests may never get through at all.
dose_main (or communication) needs a mechanism to treat all types of data reasonably fairly. Some kind of round-robin between different data type identifiers and also pd and normal traffic (which I think currently use the same data type identifier).
Pd and normal traffic may need to have different data type identifiers, if the data type identifiers is the basis of how we solve this issue.
It is also not completely clear whether this mechanism belongs in communication or in dose_main. We want to avoid adding latency to the data transmission.
The text was updated successfully, but these errors were encountered: