DATA DISTRIBUTION SERVER EVALUATION
With a “hub-and-spoke” system consisting of a Web server or a centralized database, client-server will work well for the application. If multiple nodes are generating information, client-server architectures will require that all information be sent to the server for subsequent redistribution to the clients. TCP also requires dedicated resources for each connection and does not scale well for extended data distribution in larger systems due to the set-up time and maintenance needs of each connection.
Message Passing Message-passing architectures implement queues of messages as a fundamental design paradigm. Message passing allows direct peer-to-peer connections. A message-passing design is best if you do not need a data-centric model: With this architecture there is no real model of the data itself, only a model of a means to transfer data. Application portability and integration with nodes outside the OS are an issue to consider.
Publish-Subscribe Adds a data model to messaging such that messages pass directly between the communicating nodes without requiring intermediate servers. Multiple sources and sinks are defined within the model for natural redundancy and fault tolerance. If the message-passing dataflow was difficult to draw, try it again with each node just publishing the data it knows and subscribing to what it needs. The data model means you can essentially ignore the complexity of the data flow.
- Low Latency
- Ultra Low Latency
- In-memory database
- Supports just-in-time data delivery
- No distribution hub required
- Data serialization
- Data routing and distribution rules definable ?
- Data Object Classification and Taxonomy
- Supports access control lists
- Supports certified delivery
- Object Management Group (OMG) specification
- Data Distribution Service (DDS) Interoperability
- Real-Time Publish Subscribe (RTPS)