Distributed Systems types of distributed systems Prof Leonardo

Distributed Systems – types of distributed systems Prof. Leonardo Mostarda University of Camerino Prof. Leonardo Mostarda-- Camerino, 1

Outline z. Types of distributed systems z. Distributed computing systems y. Cluster computing y. Grid computing z. Distributed information systems y. Transaction processing systems y. Enterprise integration applications z Distributed pervasive systems: Home systems, Health care systems, Wireless sensor networks

Types of distributed systems Distributed Systems Distributed Computing Systems Distributed Information Systems Distributed Pervasive Systems Clustering computing Transaction Processing Systems Home systems Grid computing Enterprise Application Integration Health care Systems Sensor Networks

Distributed computing systems z Distributed computing systems are used for highperformance computing tasks. z Roughly speaking, one can make a distinction between two subgroups. y Cluster computing y Grid computing

Cluster Computing Systems z Connection of relatively simple computers in a high-speed network to do parallel programming y A program is running on multiple machine z Linux-based Beowulf clusters y A master that allocates programs y Libraries are used for executing parallel programs they provide advance message-based communication facilities Linux-based Beowulf clusters

Cluster Computing Systems z IBM cluster computing y 524, 288 POWER 7 cores y 24 TB ram y fast interconnection: 1000 Gbsec z European Centre for Medium-Range Weather Forecasts Linux-based Beowulf clusters

Grid Computing Systems z Grid systems are composed of heterogeneous computers in different administrative domains z A key issue is that resources from different organisations are brought together to enable collaboration in the form of Virtual organisation: y Servers, storage, sensors, telescopes z Software provides access to different resources residing in different administrative domains

Grid Computing Systems z Real example of grid computing? Screensaver Lifesaver Centre for Computational Drug Discovery's project SETI (Search for Extraterrestrial Intelligence) @Home project

Cluster VS grid z When would you choose a cluster? Any guess? z It is about the amount of data you need to move between computers? Prof. Leonardo Mostarda-- Camerino, 9

Types of distributed systems Distributed Systems Distributed Computing Systems Clustering computing Grid computing Distributed Information Systems Transaction Processing Systems Enterprise Application Integration Distributed Pervasive Systems Home systems Health care Systems Sensor Networks

Distributed Information Systems z “A collection of information systems that cooperate appearing a single information system” z Several ways in which the integration can take place y Client sending a request to a centralised DBMS y Client performing a distributed transaction y Applications directly communicate with each other z We discuss two types of distributed information systems y Transaction processing systems y Enterprise application integration

Transaction Processing Systems (1) z Operations on a database are usually executed in the form of transactions; y ATM transactions, Supermarket payments, ebay z A transaction can contain: y WRITE, READ, RPC calls (transactional RPCs) z. A transaction is enclosed between BEGIN_TRANSACTION and END_TRANSACTION

Transaction Processing Systems (2) z Characteristic properties of transactions: y Atomic: To the outside world, the transaction happens indivisibly. y Consistent: The transaction does not violate system invariants. y Isolated: Concurrent transactions do not interfere with each other. y Durable: Once a transaction commits, the changes are permanent.

Transaction Processing Systems (3) Are transaction always on a single database? A nested transaction is composed of several subtransactions They improve: Performance

Transaction Processing Systems (4) What does it happen when the sub transaction commits and the transaction abort?

Early enterprise middleware systems z TP monitor handled distributed (or nested) transactions forming the core for integrating applications at the server or database level. z Its main task was to allow an application to access multiple server/databases by offering it a transactional programming model. Prof. Leonardo Mostarda-- Camerino, 16

Enterprise applications z Applications became decoupled from the databases they were built on. z This created the need of integrating applications z Applications should communicate by each other and not merely by means of the request/reply behaviour supported by transaction processing systems z Applications can directly exchange information as shown in the following

Enterprise Application Integration Middleware as a communication facilitator in enterprise application integration.

Enterprise Application Integration z Several types of communication middleware exist y Remote Procedure Calls (RPC): an application sends a request to a remote application by means of a local procedure call y Remote method invocation (RMI): is the same as RPC except that it operates on objects What is the main disadvantage of RMI and RPC? Caller and callee need to be up and running They need to know how to refer to each other. The solution is message oriented middleware (pub/sub)

Types of distributed systems Distributed Systems Distributed Computing Systems Clustering computing Grid computing Distributed Information Systems Distributed Pervasive Systems Transaction Processing Systems Home systems Enterprise Application Integration Health care Systems Sensor Networks

Distributed Pervasive Systems z Information processing is integrated into everyday objects and activities. z someone "using" pervasive computing engages many computational devices and systems simultaneously, and may not necessarily even be aware that they are doing so. z Nodes are very different from traditional desktops since they are y mobile y can be embedded y Battery powered y Equipped with wireless communication

Distributed Pervasive Systems z A distributed pervasive system is part of our surrounding environment z Devices need to discover the services and adapt to environment z Some examples y Home systems y Electronic Health Care Systems y Wireless sensor networks

Smart home Self Managing Self configuring Personal Space

Electronic Health Care Systems z (a) a local hub is managing the BAN and collecting data. This is offloaded to a centralised repository z (b) a continuous wireless connection. z Unlike home systems, health care systems cannot move to a singleserver systems ( this results in devices with minimal functionality). On the contrary: they need in-network data processing

Wireless sensor networks z WSNs consists of spatially distributed autonomous sensors that cooperatively monitoring physical or environmental conditions. Unique characteristics limited processing capability small amount of memory limited power source.

Wireless sensor networks WSN`s present a lot of challenges z z z Energy consumption Programming Security Routing Software engineering Middleware

Sensor Networks (2) z One would like to query the sensor network to obtain info z We can organise the sensor network as a distributed database z (a) sensors do not cooperate but forward data

Sensor Networks (3) z (b) Relevant sensors compute an answer

Sensor Networks (3) z The solutions presented are not very attractive: y (a) Wastage of energy y (b) Less data to be returned to the operator z We need in-network processing capabilities z A good solution is pub/sub systems z Let us see an example of a pub/sub infrastructure

Sensor Networks (3) z Components (e. g. , C 1, C 2) provide/use readings such as temperature and humidity z The policy manager performs some computation and decides weather it needs to publish readings or subscribes to new readings z As you can see the computation is performed by the nodes that is in-network computation Picture from the paper “A Policy-based Publish/Subscribe Middleware for Sense-and. React Applications” By Leonardo Mostarda, Giovanni Russello and Naranker Dulay

Summary z. Types of distributed systems z. Distributed computing systems y. Cluster computing y. Grid computing z. Distributed information systems y. Transaction processing systems y. Enterprise integration application z Distributed pervasive systems: Home systems, Health care systems, Wireless sensor networks

15 minutes pause Prof. Leonardo Mostarda-- Camerino, 32

Questions? Prof. Leonardo Mostarda-- Camerino, 33