In case a single client (i.e. libpaf context) has a large number of concurrent protocol transactions toward a server, processing latency per transaction increases.
This may happen in case a client decides to publish a very large number of services at the same time, or in case it has a very large number of subscriptions. "Large" here need to be something in the thousands or tens of thousands to become noticeable.
This slow-down is primarily due to the fact that ongoing transactions are stored in a linked list, and finding a particular transaction requires O(N) time.
In case a single client (i.e. libpaf context) has a large number of concurrent protocol transactions toward a server, processing latency per transaction increases.
This may happen in case a client decides to publish a very large number of services at the same time, or in case it has a very large number of subscriptions. "Large" here need to be something in the thousands or tens of thousands to become noticeable.
This slow-down is primarily due to the fact that ongoing transactions are stored in a linked list, and finding a particular transaction requires O(N) time.