This is an old revision of the document!
====== OMG API4KB Distributed Architecture Styles ====== ===== Introduction: Software Communication ===== The two things needed for software communication are a message transport and a message protocol. A message transport is the physical means to convey a message. Examples of transport mechanisms include Ethernet, shared memory, and the Peripheral Component Interconnect (PCI) bus. Wi-Fi and Bluetooth are examples of wireless transports. If the source and destination happen to be in the same virtual address space; the transport can be optimized into a set of parameters that are passed in a simple function call. Otherwise communication must take place through a transport facility. Virtually all CORBA implementations automatically recognize when this optimization can be used. All physical message transports including Ethernet and Wi-Fi just transport bytes. It is left up to the source and destination to agree on a message protocol. The protocol specifies the set of rules for how these messages will be serialized for transport. A protocol could be ad-hoc and unsophisticated or it might have the sophistication necessary for the source and destination to remain oblivious of the distributed nature of the system. This in turn allows the developer to choose the style that best fits the application. The API4KB RFP indicates that we must support the following architecture styles (although it uses the term platform mappings): - Mandatory: A Java interface - Mandatory: WSDL - Mandatory: REST - Optional: OMG IDL Both CORBA and Java RMI serialize messages in compliance with the Internet Inter-ORB Protocol (IIOP) standard. This is a comprehensive standard, which is usually implemented with support from automated tools. On the other hand, web services only specifies the use of XML and leaves up to the software developer(s) to ensure that the source and the destination have the same understanding of the XML content. There are many tools to support this understanding. WSDL specifies that messages will comply to an XML schema. If WSDL is not being used and no XML schema has been specified; the source and destination must at lease have some sort of mutual understanding of an implied schema or else they run the risk that parts of their messages will be lost or misunderstood. In the following three categories of communication paradigms are distinguish according to their decreasing strength of API4KB coupling. ===== Direct Strongly-Coupled API4KB Access ===== Strong coupling with the local client requiring direct knowledge of the (downloaded) API4KB **Artifacts** or direct knowledge about how the inter-process interaction and access with the remote API4KB works in ad-hoc network programming (e.g. via socket programming). Example: [[http://www.corporate-semantic-web.de/ontomaven.html|OntoMaven and RuleMaven]]. ===== Loosely-coupled Remote Invocation via API4KB Interfaces ===== Wide range of techniques based on a loosely-coupled two-way exchange via an interface between communicating entities. ==== Request-Reply Protocols ==== Protocols involve pairwise exchange of messages from client to server and from server back to client with the first message containing an encoding of operation to be executed at the server, the second message contains the result (encoded as an array of bytes). Paradigm is rather primitive (in contrast to RPC/RMI) and typically only used for e.g. embedded systems where performance is very important. Approach is also used by e.g. the http protocol. ==== Remote Procedure Calls ==== Examples, e.g. Web Services, stateless REST Web Services, Enterprise Service Bus, ... ==== Remote Method Invocation ==== Resembles RPC in the world of distributed objects. **Distributed Object Middleware**, e.g. Java RMI (which is restricted to Java). OMG CORBA is a multi-language solution with a declarative Interface Description Language (IDL). Usually developers choose to invoke CORBA methods through a static interface, which is obtained by using an automated tool to translate the IDL into the chosen implementation language. However, it is also possible to formulate a CORBA message using the facilities of the Dynamic Invocation Interface (DII). ==== Distributed Components ==== A unit of composition with contractually specified interfaces and explicit content dependencies only. Component is specified in terms of a contract which includes a set of provided interfaces (interfaces that the component offers as a service to other components) and required interfaces (dependencies that this component has on other components). Container provides managed server-side hosting environment for components and deals with the distributed systems and middleware issues. Examples, e.g. Java Beans, Corba Component Model, [[http://www.corporate-semantic-web.de/ontomaven.html|OntoMaven Aspect-Oriented Component Model]], ... ===== Decoupled Indirect Communication ===== Indirect decoupled techniques where sender and receiver are time and space uncoupled via an intermediary. Indirect communication between entities in a distributed system through an intermediary with no direct coupling between the sender and the receiver(s), e.g. event routing in publish-subscribe middleware (based e.g. on peer-to-peer), streaming to a cloud, .... ==== Publish Subscribe and Distributed Event Based Systems ==== Publish-subscribe with event-based communication through propagation of events (via an underlying overlay network, e.g. structured and unstructured peer-to-peer or other broker overlay). Publishers publish structured events to an event service (responsible for event routing and matching) and subscribers express interest in particular events through subscriptions. Distributed Event Based Systems and Event Streaming with Complex Event Processing (CEP) in Event Processing Agents (EPAs) deployed in Event Processing Networks (EPNs). ==== Group communication ==== Broadcast and multicast. ==== Shared Resources ==== Examples, e.g. Tuple Spaces, Distributed Shared Memory, Message Queues (e.g., JMS, Active MQ, ...) ==== Asynchronous Messaging Libraries ==== Libraries exist which combine and can be used according to a number of the different, above mentioned messaging patterns, e.g. * [[http://en.wikipedia.org/wiki/%C3%98MQ|0MQ]] (ZeroMQ) is a lightweight messaging system specially designed for high throughput and low latency scenarios as in IoT, low-level event streaming etc. It provides "sockets" (a many-to-many connection generalizing the concept of network socket) which each operate according to a specific messaging pattern (e.g. RPC, pub-sub, ...), which (in contrast to e.g. more advanced messaging queue servers and enterprise service bus middlewares) need to be manually implement by combining various pieces of the framework (see ad-hoc network programming with sockets and devices in strong coupling category). * higher-level message queue middleware such as Erlang (RabbitMQ), C (beanstalkd), Ruby (Starling or Sparrow), Scala (Kestrel, Kafka) or Java (ActiveMQ), and Enterprise Service Bus middleware such as OpenESB, Mule ESB, ... Examples, e.g. [[http://responder.ruleml.org|RuleML RuleResponder]] is implemented as a Staged Event-Driven Architecture (SEDA) using the Mule ESB and ActiveMQ or RabbitMQ. It can use all kinds of synchronous and asynchronous transport protocols including e.g. AMQP, JMS, HTTP Rest, SOAP, etc.. ===== Summary of Dimensions ===== * Time Uncoupling - The sender and the receiver(s) can have **independent lifetimes** * Space Uncoupling - The sender** does not know or need to know the identity of the receiver(s)**, and vice versa * Synchronous vs. Asynchronous Communication - sender sends a message and then continues **without blocking** * Centralized vs. Distributed Architecture (+hybrid) * Structured vs. Unstructured Topology * Placement: Multiple Servers, Proxy/Cache, Mobile Code * Architecture Styles: Client-Server, Broker Overlay, Peer-to-Peer * Multiplicity of Message Recipients (Group Communication, Brokered/Brokerless) * Directionality of Communication (Uni-/Bi-directional) * Communication Entities: Processes, Objects, Components, Services / Agents * Classification of Endpoints (Single-/Multi-Sorted)