Implementing a transactional topology
In the previous chapters, we have concerned ourselves with data that was not "transactional" in nature, not in the way we typically think of things such as financial transactions. As a result, one can potentially tolerate system failure because there is no direct monetary implication to each transaction that may be affected by any failure cases, especially given that transactional schematics come with a cost, both in performance and storage. The recipes in this chapter deal with scoring transactions that require transactional schematics, and it is therefore relevant to understand how to achieve exactly-once schematics with Storm at this stage.
The transactional schematics of Storm, like most aspects of Storm, ...