Distributed Operating System Concepts PDF

Operating System concept Distributed Operating Systems  Distributed Systems What is a distributed system? “A collection of independent computers that appears to its users as a single coherent system.” Characteristics – Prone to failure – Limited bandwidth – Latency ...

Operating System concept Distributed Operating Systems  Distributed Systems What is a distributed system? “A collection of independent computers that appears to its users as a single coherent system.” Characteristics – Prone to failure – Limited bandwidth – Latency  Definition of a DS (2) A distributed system organized as middleware.Note that the middleware layer extends over multiple machines – To support heterogeneity and at the same time offering a single-system view – Provides means for components to communicate and at the same time hides the differences in OS and hardware from each application  Characteristics of a DS Differences between the computers (platforms, HW, SW) and the ways in which they communicate are hidden from users – Heterogeneous computer and networks Easy to expand or scale Continuously available (yet some parts may temporarily out of order and) – users should not notice that parts are being replaced or fixed , or that new parts are being added  Goals Of DS – Easily Connect Users / Resources. – Exhibit Transparency. – Support Openness. – Be Scalable – In size. – Geographically. – Administratively.  Goals Of Distributed Systems (Transparency) A DS is transparent if: It presents itself to users as if it were a single system. – Hide the fact that its processes and resources are physically distributed across multiple computers Transparency Types in a DS Transparency Types in a DS (contd)  Goals Of Distributed Systems (Openness) Goals Of Distributed Systems (Openness) Offers services according to standard rules with clear syntax protocols of computer networks – Interoperability Two implementations of systems from different manufacturers can co-exist and work together – Portability Applications for one DS can work on another DS without modification – extensibility Easy to configure, add, delete, replace components without affecting other parts of the system  Goals Of Distributed Systems:(Scalability) Three dimensions of scalability problems: – Size: Number of users and/or processes (easy to add) – Geographical: Maximum distance between nodes (resources may be far apart) – Administrative: Number of administrative domains (easy to manage) Performance must not degrade with a growing DS, – however, in practice a system exhibits some loss of performance as the system scales up Avoid any form of a centralized solution within a DS, because each central “resource” might become a bottleneck: – Components /services (e.g. Single Server) or – Date /Tables (e.g. directories in DFS) – Algorithms (e.g. deadlock detection) Scalability Problems: (Size Scalability)  Techniques for Scaling In general, scalability problems appear in the form of performance degradation caused by limited capacity of servers and networks. 3 techniques for scaling – Hiding communication latencies: – Distribution: Partition data and computations across multiple machines in the system: Decentralized naming services (e.g., DNS) Decentralized information systems (e.g., WWW) – Replication: Make copies of data available at different machines (e.g., replicated file servers, databases, mirrored websites, etc) Increase availability and balance load Caching : Allow client processes to access local copies – Web caches (e.g., browser/web proxy) – File caches (e.g., server or client) Consistency problems  Hiding communication latencies To achieve geographical scalability Simply by avoid waiting for responses. – Asynchronous communication,  Scaling Techniques Distribution: Partition data and computations across multiple machines in the system: Decentralized naming services (e.g., DNS) Decentralized information systems (e.g., WWW)  Scaling Techniques Replication: Make copies of data available at different machines (e.g., replicated file servers, databases, mirrored websites, etc) – Increase availability and balance load which results in better performance – Hide some communication latency by having a near by copy. Caching : Allow client processes to access local copies – Similar to replication, it makes a copy of a recourse in the proximity of the client. However, the decision of caching is made by the client not the owner of the resource – Also, caching happens on demand while replication is often planned in advance  Scaling Techniques May create consistency problems: modifying one copy makes that copy different from others Toleration of inconsistency depends of the resource usage: – Weak consistency – Web caches (e.g., browser/web proxy): – File caches (e.g., server or client) – Strong consistency – Stock exchange  Summary: Scaling Techniques 1.Hiding Communication Latencies – The web-server example (cache and do at all once) 2.Distribution – Split a component and spread the parts around – E.g., Domain Name System (DNS) 3.Replication – Replicate data around