Message size versus Message tree

When the size of messages that are being processed is being discussed, then the terms small and large are often used. It is difficult to quantify these terms in a generic way even to explain where the crossover point is from small to large.

When messages are a couple of kilobytes in a size, then they are classed as small messages. When messages are in the tens of megabytes range, then some users would class them as large. However, these attempts at a classification are just considering the size of the bitstream data that is being processed. Nearly all transports in IBM® Integration Bus read the incoming message data into a contiguous memory buffer, and as such a message flow allocates the required storage. But the message tree is larger than just the buffer. For example, if there is a repeating element to the message flow, then the size of the input data can grow over time until it is much larger than was initially planned for.

Some of the file transports such as File and TCP/IP are able to stream data in chunks. For such transports, it is not necessary to read the whole of the input data into memory at the same time. The amount of contiguous storage that is required to read a "chunk" depends on the options that are selected on the input node.