User Interface
MODEL-VIEW-CONTROLLER (MVC)						A system may offer multiple user interfaces. Each user interface depicts all or part of some application data. Changes to the data should be automatically and flexibly reflected to all the different user interfaces. It should be also possible to easily modify any one of the user interfaces, without affecting the application logic associated with the data. The system is divided into three different parts: a Model that encapsulates some application data and the logic that manipulates that data, independently of the user interfaces; one or multiple Views that display a specific portion of the data to the user; a Controller associated with each View that receives user input and translates it into a request to the Model. Views and Controllers constitute the user interface. The users interact strictly through the Views and their Controllers, independently of the Model, which in turn notifies all different user interfaces about updates.
PRESENTATION-ABSTRACTION-CONTROL (PAC)						An interactive system may offer multiple diverse functionalities that need to be presented to the user through a coherent user interface. The various functionalities may require their own custom user interface, and they need to communicate with other functionalities in order to achieve a greater goal. The users need not perceive this diversity but should interact with a simple and consistent interface. The system is decomposed into a tree-like hierarchy of agents: the leaves of the tree are agents that are responsible for specific functionalities, usually assigned to a specific user interface; at the middle layers there are agents that combine the functionalities of related lower-level agents to offer greater services; at the top of the tree, there is only one agent that orchestrates the middle-layer agents to offer the collective functionality. Each agent is comprised of three parts: a Presentation takes care of the user interface; an Abstraction maintains application data and the logic that modifies it; a Control intermediates between the Presentation and the Abstraction and handles all communication with the Controls of other Agents.
Components and Connectors (C2)						An interactive system is comprised of multiple components such as GUI widgets, conceptual models of those widgets at various levels, data structures, renderers, and of course application logic. The system may need to support several requirements such as: different implementation language of components, different GUI frameworks reused, distribution in a heterogeneous network, concurrent interaction of components without shared address spaces, run-time reconfiguration, multi-user interaction. Yet the system needs to be designed to achieve separation of concerns and satisfy its performance constraints. The system is decomposed into a top-to-bottom hierarchy of concurrent componentsthat interactasynchronouslyby sending messages throughexplicit connectors. Components submit request messages upwards in the hierarchy, knowing the components above, but they send notification messages downwards in the hierarchy, without knowing the components lying beneath. Components are only connected with connectors, but connectors may be connected to both components and other connectors. The purposes of connectors is to broadcast, route, and filter messages.
Hierarchical Structure
CLIENT-SERVER						Two components need to communicate, and they are independent of each other, even running in different processes or being distributed in different machines. The two components are not equal peers communicating with each other, but one of them is initiating the communication, asking for a service that the other provides. Furthermore, multiple components might request the same service provided by a single component. Thus, the component providing a service must be able to cope with numerous requests at any time, i.e. the component must scale well). On the other hand, the requesting components using one and the same service might deal differently with the results. This asymmetry between the components should be reflected in the architecture for the optimization of quality attributes such as performance, shared use of resources, and memory consumption. The CLIENT-SERVER pattern distinguishes two kinds of components: clients and servers. The client requests information or services from a server. To do so it needs to know how to access the server, that is, it requires an ID or an address of the server and of course the server’s interface. The server responds to the requests of the client, and processes each client request on its own. It does not know about the ID or address of the client before the interaction takes place. Clients are optimized for their application task, whereas servers are optimized for serving multiple clients.
PEER-TO-PEER						Consider a situation similar to that of a CLIENT-SERVER, but in contrast to CLIENT-SERVER, there is no distinction between the components: each component might both provide services and consume services. When a component provides a service it must perform well according to the demands of the requesting components. Each component must know how to access other components. In the PEER-TO-PEER pattern each component has equal responsibilities, in particular it may act both as a client and as a server. Each component offers its own services (or data) and is able to access the services in other components. The PEER-TO-PEER network consists of a dynamic number of components. A PEER-TO-PEER component knows how to access the network. Before a component can join a network, it must get an initial reference to this network. This is solved by a bootstrapping mechanism, such as providing public lists of dedicated peers or broadcast messages (using IMPLICIT INVOCATION) in the network announcing peers.
MASTER-SLAVE						The Master-slave pattern supportsfault toleranceandparallel computation. The master component distributes the work among identical slave components, and computes a final result from the results the slaves return. The Master-slave pattern is applied for instance in process control, in embedded systems, in large-scale parallel computations, and in fault-tolerant systems. An application area for the Master-slave pattern is fault tolerance: the master delegates the job to be done to several slaves, receives their results, and applies a strategy to decide which result to return to the client. One possible strategy is to choose the result from the first slave that terminates. Another strategy is to choose the result that the majority of slaves have computed. This is fail-proof with respect to slaves (the master can provide a valid result as long as not all slaves fail), but not with respect to the master. Failure of slaves may be detected by the master using time-outs. Failure of the master means the system as a whole fails. Another application area is parallel computing: the master divides a complex task into a number of identical subtasks. An example is matrix computation: each row in the product matrix can be computed by a separate slave. A third application area is that of computational accuracy. The execution of a service is delegated to different slaves, with at least three different implementations. The master waits for the results, and applies a strategy for choosing the best result (for instance the average, or the majority). The Master-slave pattern is an example of the divide-and-conquer principle. In this pattern, the aspect of coordination is separated from the actual work: concerns are separated. The slaves are isolated: there is no shared state. They operate in parallel. The latency in the master-slave communication can be an issue, for instance in real-time systems: master and slaves live in different processes. The pattern can only be applied to a problem that is decomposable.
Message Passing Mechanism
PUBLISH-SUBSCRIBE						A component should be accessed or informed of a specific runtime event. Events are of different nature than direct interactions as in CLIENT-SERVER or PEER-TO-PEER. Sometimes a number of components should be actively informed (an announcement or broadcast), in other cases only one specific component is interested in the event. In contrast to EXPLICIT INVOCATIONS, event producers and consumers need to be decoupled for a number of reasons: to support locality of changes; to locate them in different processes or machines; to allow for an arbitrary time period between the announcement of interest in an event, and the actual triggering of the event. Still, there must be a way to inform the interested components? PUBLISH-SUBSCRIBE allows event consumers (subscribers) to register for specific events, and event producers to publish (raise) specific events that reach a specified number of consumers. The PUBLISH-SUBSCRIBE mechanism is triggered by the event producers and automatically executes a callback-operation to the event consumers. The mechanism thus takes care of decoupling producers and consumers by transmitting events between them.
BROKER						Distributed software system developers face many challenges that do not arise in single-process software. One is the communication across unreliable networks. Others are the integration of heterogeneous components into coherent applications, as well as the efficient use of networking resources. If developers of distributed systems must overcome all these challenges within their application code, they may lose their primary focus: to develop applications that efficiently tackle their domain-specific problems. A BROKER separates the communication functionality of a distributed system from its application functionality. The BROKER hides and mediates all communication between the objects or components of a system. A BROKER consists of a client-side REQUESTOR to construct and forward invocations, as well as a server-side INVOKER that is responsible for invoking the operations of the target remote object. A MARSHALLER on each side of the communication path handles the transformation of requests and replies from programminglanguage native data types into a byte array that can be sent over the transmission medium.
REMOTE PROCEDURE CALLS (RPC)						Consider the case where you want to realize an EXPLICIT INVOCATION in a distributed setting. The use of low-level network protocols requires developers to invoke the send and receive operations of the respective network protocol implementations. This is undesirable because the network access code cannot be reused, low-level details are not hidden, and thus solutions are hard to maintain and understand. REMOTE PROCEDURE CALLS extend the well-known procedure call abstraction to distributed systems. They aim at letting a remote procedure invocation behave as if it were a local invocation. Programs are allowed to invoke procedures (or operations) in a different process and/or on a remote machine.
MESSAGE QUEUING						Consider a situation similar to that of REMOTE PROCEDURE CALLS, but it is necessary to decouple the sender from the receiver to realize queuing of invocations. Queuing is necessary, for instance, when temporal outages of the receiver should be tolerated or when heterogeneous systems should be integrated. For instance, when a legacy system using BATCH SEQUENTIAL should be integrated into a distributed system, only one invocation can be handled at a time by that system. Somewhere additional messages must be queued until the system is ready to process the next message. Messages are not passed from client to server application directly, but through intermediate message queues that store and forward the messages. This has a number of consequences: senders and receivers of messages are decoupled, so they do not need to know each other’s location (perhaps not even the identity). A sender just puts messages into a particular queue and does not necessarily know who consumes the messages. For example, a message might be consumed by more than one receiver. Receivers consume messages by monitoring queues.
IMPLICIT INVOCATION (EVENT-BASED)						Consider the case where an invocation is needed, such as in EXPLICIT INVOCATION. Furthermore, the client must be decoupled in various ways from the supplier, during the delivery of the invocation and of the result: the client might not know which supplier serves the invocation; or the client may not initiate the invocation itself but is merely interested in the invocation result; or the client does not need the result right away so it can be occupied with another task in the meantime; or the supplier might not be ready to reply to the client until some condition has been met; or clients may be added or removed dynamically during the system runtime; or the client does not know that the supplier is up and running and, if the supplier is down, the system should suspend the invocation until the supplier is up again; or the client and the supplier are part of dissimilar systems and thus the invocation must be transformed, queued, or otherwise manipulated during delivery. How can such additional requirements during delivery be met? In the IMPLICIT INVOCATION pattern the invocation is not performed explicitly from client to supplier, but indirectly and rather randomly through a special mechanism such as PUBLISHSUBSCRIBE, MESSAGE QUEUING, or broadcast, that decouples clients from suppliers. All additional requirements for invocation delivery are handled by the IMPLICIT INVOCATION mechanism during the delivery of the invocation.
EXPLICIT INVOCATION						Consider a component, the client, which needs to invoke a service defined in another component, the supplier. Coupling the client with the supplier in various ways is not only harmless but often desirable. For example the client must know the exact network location of the component which offers the service in order to improve performance; or the client must always initiate the invocation itself; or the client must block, waiting for the result of the invocation, before proceeding with its business; or the topology of the interacting clients and suppliers is known beforehand and must remain fixed. How can these two components interact? An EXPLICIT INVOCATION allows a client to invoke services on a supplier, by coupling them in various respects. The decisions that concern the coupling (e.g. network location of the supplier) are known at design-time. The client provides these design decisions together with the service name and parameters to the EXPLICIT INVOCATION mechanism, when initiating the invocation. The EXPLICIT INVOCATION mechanism performs the invocations and delivers the result to the client as soon as it is computed. The EXPLICIT INVOCATION mechanism may be part of the client and the server or may exist as an independent component.
Data Accessor
SHARED REPOSITORY						Data needs to be shared between components. In sequential architectures like LAYERS or PIPES AND FILTERS the only way to share data between the components (layers or filters) is to pass the information along with the invocation, which might be inefficient for large data sets. Also it might be inefficient, if the shared information varies from invocation to invocation because the components’ interfaces must be prepared to transmit various kinds of data. Finally the long-term persistence of the data requires a centralized data management. In the SHARED REPOSITORY pattern one component of the system is used as a central data store, accessed by all other independent components. This SHARED REPOSITORY offers suitable means for accessing the data, for instance, a query API or language. The SHARED REPOSITORY must be scalable to meet the clients’ requirements, and it must ensure data consistency. It must handle problems of resource contention, for example by locking accessed data. The SHARED REPOSITORY might also introduce transaction mechanisms.
ACTIVE REPOSITORY						A system needs to have a SHARED REPOSITORY, but it should not just be passively accessed by accessor components. Clients need to be immediately informed of specific events in the shared repository, such as changes of data or access of data. “Polling” (i.e. querying in frequent intervals) the SHARED REPOSITORY for such events does not work, for instance, because this does not deliver timely information or inflicts overhead on the system performance. An ACTIVE REPOSITORY is a SHARED REPOSITORY that is “active” in the sense that it informs a number of subscribers of specific events that happen in the shared repository. The ACTIVE REPOSITORY maintains a registry of clients and informs them through appropriate notification mechanisms.
BLACKBOARD						Consider the case where a SHARED REPOSITORY is needed for the shared data of a computation, but no deterministic solution strategies are known. Examples are image recognition or speech recognition applications. However, it should be possible to realize a solution for these types of applications. In a BLACKBOARD architecture the complex task is divided into smaller sub-tasks for which deterministic solutions are known. The BLACKBOARD is a SHARED REPOSITORY that uses the results of its clients for heuristic computation and step-wise improvement of the solution. Each client can access the BLACKBOARD to see if new inputs are presented for further processing and to deliver results after processing. A control component monitors the blackboard and coordinates the clients according to the state of the blackboard.
System organization
LAYERS						Consider a system in which high-level components depend on low-level components to perform their functionality, which further depend on even lower-level components and so on. Decoupling the components in a vertical manner is crucial in order to support modifiability, portability, and reusability. On the other hand components also require some horizontal structuring that is orthogonal to their vertical subdivision. To achieve these goals, the system is structured into LAYERS so that each layer provides a set of services to the layer above and uses the services of the layer below. Within each LAYER all constituent components work at the same level of abstraction and can interact through connectors. Between two adjacent layers a clearly defined interface is provided. In the pure form of the pattern, layers should not be by-passed: higher-level layers access lower-level layers only through the layer beneath.
BATCH SEQUENTIAL						Consider a complex task that can be sub-divided into a number of smaller tasks, which can be defined as a series of independent computations. This should not be realized by one monolithic component because this component would be overly complex, and it would hinder modifiability and reusability. In a BATCH SEQUENTIAL architecture the whole task is sub-divided into small processing steps, which are realized as separate, independent components. Each step runs to completion and then calls the next sequential step until the whole task is fulfilled. During each step a batch of data is processed and sent as a whole to the next step.
CQRS						Command and Query Responsibility Segregation (CQRS) is a pattern that segregates the operations that read data (Queries) from the operations that update data (Commands) by using separate interfaces. This implies that the data models used for querying and updates are different. Compared to the single model of the data (from which developers build their own conceptual models) that is inherent in CRUD-based systems, the use of separate query and update models for the data in CQRS-based systems considerably simplifies design and implementation. However, one disadvantage is that, unlike CRUD designs, CQRS code cannot automatically be generated by using scaffold mechanisms. The query model for reading data and the update model for writing data may access the same physical store, perhaps by using SQL views or by generating projections on the fly. However, it is common to separate the data into different physical stores to maximize performance, scalability, and security;
SERVICE-ORIENTED (SOA)						In contrast to conventional software architectures primarily delineating the organization of a system in its (sub)systems and their interrelationships, the SOA captures a logical way of designing a software system to provide services to either end-user applications or other services distributed in a network through published and discoverable interfaces. SOA is focused on creating a design style, technology, and process framework that will allow enterprises to develop, interconnect, and maintain enterprise applications and services efficiently and costeffectively.While this objective is definitely not new, SOA seeks to eclipse previous efforts such as modular programing, code reuse, and object-oriented software development techniques. SOA requires that functions, or services, are defined by a description language and have interfaces that perform useful business processes. The fundamental intent of a service in an SOA is to represent a reusable unit of business-complete work. A service in SOA is an exposed piece of functionality with three essential properties. Firstly, an SOA-based service is self-contained, i.e., the service maintains its own state. Secondly, services are platform independent, implying that the interface contract to the service is limited to platform independent assertions. Lastly, the SOA assumes that services can be dynamically located, invoked and (re-)combined.
MICROSERVICE						A microservice is a software development technique—a variant of the service-oriented architecture (SOA) architectural style that structures an application as a collection of loosely coupled services. In a microservices architecture, services are fine-grained and the protocols are lightweight. The benefit of decomposing an application into different smaller services is that it improves modularity and makes the application easier to understand, develop, and test and more resilient to architecture erosion. It parallelizes development by enabling small autonomous teams to develop, deploy and scale their respective services independently. It also allows the architecture of an individual service to emerge through continuous refactoring. Microservices-based architectures enable continuous delivery and deployment.
COMPONENT-BASED (CB)