A programmer looking for a flexible messaging solution is faced with many questions and tradeoffs: What tool will offer the best performance? What will give the best flexibility and extensibility? Do I need a persistence layer or a cache? Depending on the specific application area and the structure of the data being transmitted, the requirements will be different: for example, an electronic options trading engine has different needs from a SaaS web application. This talk will compare a variety of contemporary messaging solutions and the benefits and tradeoffs involved: - Kafka offers a persistent datastore and a clustering solution based on a "pub-sub" (publish-subscribe) architecture; - RabbitMQ is a message broker that implements the Advanced Message Queueing Protocol; - ZeroMQ implements a brokerless architecture that provides a socket-like abstraction to enable a variety of possible message distribution patterns. And finally: what do you gain and what do you lose from rolling your own simple messaging solution? Under what circumstances is the DIY approach the best one? This talk offers a fast tour through a variety of messaging technologies and compares the pros and cons of each technology. Examples are shown in Python to favor simplicity and brevity of presentation over implementation details.