Why is having consecutive compute nodes bad practice?

The overhead in parsing the data/payload can be very high, so it is better to parse it in one conmpute node rather than repeating the parsing in several consecutive nodes. 

Thanks both. By the way, great webcast Francois!

Because you copy the entire message from the input tree to the output tree for each compute node, adding a performance overhead to the entire message flow. This is important if your message flow will be used in a high-volume scenario only.

You can avoid this by using database nodes, which does not use input and output trees, but rather allows you manipulate the message tree directly

Rgds. Anders Heuch