Distributed Systems

Distributed Systems

1. CHAPTER N-2

   • 2.1 ARCHITECTURAL STYLES
     • Desing a distributed system is very important the software architectures
       • Types
         • Layered architectures
         • Layered communication protocols
         • Application layering
           • Many people have advocated a distinc-tion between three logical levels, essentially following a layered architecturalstyle:
             • The application-interface level
             • The processing level
             • The data level
         • Object-based and service-oriented architectures
         • Resource-based architectures
         • Publish-subscribe architectures
   • 2.4 EXAMPLE ARCHITECTURES
     • In this section we pay explicit attention to organizing distributed systems as high-level feedback-control systems allowing automatic adaptations to changes.
       • The Feedback Control Model
         • Examples on how to monitor, analyze, and correct distributed systems in an automatic fashion. These examples will also illustrate this distinction between logical and physical organization
           • Example: Automatic Component Repair Management in Jade
             • Jade is built on the Fractal component model, a Java implementation of a framework that allows components to be added and removed at runtime
               • Jade uses the notion of a repair management domain.
                 • Such a domain consists of a number of nodes, where each node represents a server along with the components that are executed by that server.
                   • There is a separate node manager which is responsible for adding and removing nodes from the domain.
                     • The node manager may be replicated for assuring high availability.
                     • The approach followed by Jade is an example of self-management: upon the detection of a failure, a repair policy is automatically executed to bring the system
           • Example: Differentiating Replication Strategies in Globule
             • Gloube is a collaborative content distribution network (Pierre and van Steen, 2006). Globule relies on end-user servers being placed in the Internet, and that these servers collaborate to optimize performance through replication of Web pages
               • The origin server then looks for the nearest existing replica server that could act as edge server for that client, and subsequently computes the latency to that server along with the maximal bandwidth.
           • Example: Systems Monitoring with Astrolabe
             • is a system that can support general monitoring of very large distributed systems.
             • is divid in zones
               • Every host runs an Astrolabe process, called an agent, that collects information on the zones in which that host is contained.
                 • Combined with a few enhancements to SQL, it is not hard to imagine that more informative queries can be formulated. Queries such as these are continuously evaluated by each agent running on each host
                   • This is possible only if zone information is propagated to all nodes that comprise Astrolabe
               • The top-level zone covers all hosts
               • The lowest-level zones consist of just a single host, which are subsequently grouped into zones of increasing size.
         • There are essentially three elements that form the feedback control loop
           • The system itself needs to be monitored.
           • Analyzes the measurements
           • Compares these to reference values
   • 2.3 SYSTEM ARCHITECTURE
     • Depending on where the components are located and the way they communicate, we will be talking about centralized, decentralized and hybrid architectures.
       • Centralized organization
         • To understand and understand distributed systems a little better, it is beneficial to understand that there are two parts, the first the client and the second, the services, thanks to this it is possible to abstract circumstantial issues that do not require knowing the client.
           • This type of architecture is called centralized because the server is the one that manages the resources depending on the requests that a client makes.
             • Simple client-server architecture
               • This architecture is widely used and presents a clear distinction between the end user and the services they are going to use, it consists of two parts: the client part, usually the graphical interface part and the part of the server where the services are hosted.
                 • These Two capable usually interact through the tcp / ip protocol, which means that they are connection-oriented, to get a great benefit from this type of architecture, it is recommended to deploy them in a wide network environment and not locally, since locally there is no performance different from an application without this type of architecture.
             • Multilayer or level architectures
               • It is an evolution of the client server architecture, this evolution occurred due to multiple discussions, abstractions and management of responsibilities where it was already found that the graphic part can be deployed only in the client part, but the total complexity was assumed by the server where not only the services were hosted but also the persistence of the information.
                 • The last part being totally inherent to the server's work, then a better approach was to divide these three layers with different responsibilities to reduce complexity and favor the scaling and maintenance of the systems.
       • Decentralized Organization
         • An evolution of multilayer systems was the way to distribute applications or components within different physical devices, each operating at its own pace. two types of distribution are conceived.
           • Distribution vertically dividing the functions within different machines associated as a working group.
           • Horizontal distribution, in this type of distribution it is physically divided into different machines, either the client or the server, but these parts must be logically the same, this guarantees the complete logical operation of each physical unit.
             • Peer to peer systems (peer to peer)
               • With the horizontal distribution a new problem begins, since communication is now carried out between equal systems through tcp requests, then the way of handling and communicating is divided into two
                 • Structured equal systems
                   • Commonly each node or system has a list of neighboring systems which is known or defined as random graph
                 • Unstructured peer systems
                   • One of the problems of unstructured systems is scalability, it is very complex because there is no deterministic way of handling connections, so this concept arises where the key is to have specific nodes with information indexes called super peers, the others are called weak peers
                 • Hierarchically organized equal systems
                   • Commonly grouped by a typology that allows a control over how to communicate between systems, for example a ring or binary tree typeface.
       • Hybrid Systems
         • Edge-Server Systems
           • Commonly located on the internet, these servers serve as a bridge between the local network and internet providers in order to browse or receive content from the internet.
         • Collaborative distributed systems
           • They are systems that start with a centralized organization base such as the client server architecture, but that over time tend to adopt decentralized organization measures combining their use and having a component in charge of balancing the load and managing the different resources peer to peer. that may exist
   • 2.2. MIDDLEWARE ORGANIZATION
     • In particular,there are two important types ofdesign patternsthat are often applied to theorganization of middleware: wrappers and interceptors.
       • Wrappers
         • Is a special component that offers an interfaceacceptable to a client application, of which the functions are transformedinto those available at the component. In essence, it solves the problem ofincompatible interfaces.
       • Interceptors
         • Is nothing but a software construct that willbreak the usual flow of control and allow other (application specific) codeto be executed. Interceptors are a primary means for adapting middlewareto the specific needs of an application.
           • Such a remote-objectinvocation is carried out in three steps
             • ObjectAis offered a local interface that is exactly the same as the inter-face offered by objectB.Acalls the method available in that interface.
             • The call byAis transformed into a generic object invocation, madepossible through a general object-invocation interface offered by themiddleware at the machine whereAresides.
             • Finally, the generic object invocation is transformed into a message thatis sent through the transport-level network interface as offered byA’slocal operating system.
     • Modifiable middleware
       • he need for adaptation comes from the fact that the environmentin which distributed applications are executed changes continuously.
         • Changesinclude those resulting from mobility, a strong variance in the quality-of-service of networks, failing hardware, and battery drainage, amongst others.
           • A system may either be configured statically at design time,or dynamically at runtime.

2. IMPORTANT DETAILS

   • The organization of distributed systems is mostly about the software components that constitute the system. These software architectures tell us how the various software components are to be organized and how they should interact
   • An important goal of distributed systems is to separate applications from underlying platforms by providing a middleware layer
     • The middleware its main purpose is to provide distribution transparency
   • Autonomic systems: system monitor its own behavior and taking appropriate measures when needed.