Publish-Subscribe Overview

An RTserver process routes messages between RTclient processes. A key feature of SmartSockets is the ability to distribute RTserver and RTclient processes over a network. Different processes can be run on different computers, taking advantage of all the computing power a network has to offer. RTservers and RTclients can be dynamically started and stopped while the system is running.

The feature set of SmartSockets publish-subscribe architecture is layered on top of connections and messages, but adds greater function and ease of use. Some of these functions are listed below.

RTserver and RTclient have simplified setup and control through options, which require no programming.

RTserver can partition a group of RTclients into a project.

RTclient and RTserver use logical addresses called subjects for the sender property and destination property of messages, which enable powerful yet simple publish-subscribe services.

Subjects are arranged in a hierarchical namespace with wildcard capabilities, which makes it easier to build large projects.

A group of RTservers can distribute the load of publish-subscribe message routing with dynamic message routing that offers greater scalability and flexibility.

Routing between RTservers can be further optimized to use the lowest cost route by assigning a cost to each path between RTservers.

Additional callbacks are available including subject, server create, and server destroy.

An RTclient can automatically start an RTserver, automatically restart an RTserver, and even continue running when an RTserver is temporarily unavailable.

RTserver and RTclient have advanced guaranteed message delivery (GMD) capabilities, such as automatic recovery from most network failures.

Much information about RTservers and RTclients can be monitored, which allows you to easily debug, examine, and control your projects.

Load balancing can be used to publish messages to one subscriber instead of all interested subscribers.

The RTclient API can be used safely in multithreaded programs.

This chapter describes RTserver and RTclient composition, how to work with the RTclient Application Programming Interface (API), and how to work with RTserver. Monitoring is mentioned briefly here. See Chapter 5, Project Monitoring, for a full discussion on monitoring.

Most RTclients have a single, global connection to one RTserver. Generally, this is the only connection required, and this connection forms the basis for most RTclient-RTserver interactions. The focus of this chapter is on understanding this connection and the interactions involved. You should be thoroughly familiar with this connection and the relevant APIs before delving into more complex connections. Connecting to an RTgms process instead of an RTserver process for the single, global connection requires an understanding of multicasting and a special set of PGM options. For more information, see TIBCO SmartSockets Multicast. It is also possible to connect to multiple RTservers, using a special multiple connection instead of the global connection. However, this type of connection requires a completely different set of APIs, and each individual connection must be created and configured discretely. For more information, see Connecting to Multiple RTservers.