The Cascade DataHub buffers data that will be transmitted to a client such that if it knows that more incoming data is available, it will hold off outgoing transmissions until it has a complete data set to send onward to the client. This does not introduce extra latency because the DataHub will only accumulate data destined for a client that arrives together in an incoming message.

This buffering greatly increases efficiency by reducing thread context switching and by giving its protocol-specific data transmitters an opportunity to collect more than one data change into a single outgoing message.

One of the side-effects of buffering is that the old value queue will be more likely to fill. If your application is very sensitive to every change of value, then it may be necessary to turn off the buffering. The CPU penalty for turning off buffering is very high, perhaps as much as 200% in heavy load conditions.

How to Optimize

Do not use un-buffered data delivery in high load conditions unless you absolutely must. Please refer to How to Optimize, Old value queueing and un-buffered delivery for details.