Goscinski, A. (1991) Distributed Operating Systems: The Logical Design. Addison-Wesley Publishing Company, Reading, MA.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....concept. Names are used for identifying objects, be it files, services, processors or users. When a user looks for an object, he will need to know who controls the object. This information will often be unknown. Names are, however, primarily used for exchanging messages between programs. In [22], names are observed as often referring to services, abstract entities that respond to requests of a certain type. Needham [41] goes a step further, and recognizes the entire name to basically be a string of symbols identifying an object. He identifies the string to consist of three parts: The ....

....system s ability to scale. With the growth of a distributed system comes an increase in the number of objects that need to be named. The system must be able to cope with a dynamically increasing amount of objects. To achieve all this, some kind of a name service must be present in the system. In [22], this service is defined as a facility which enables clients to name objects and subsequently use these names to refer to those objects. The facility realizing this is called a name server. Upon request, it will provide a client process with both the physical address and the route to any object ....

[Article contains additional citation context not shown here]

Goscinski, Andrzei: "Distributed operating systems. The logical design", Addison-Wesley, 1991.

....area tree, and the addresses, on the fixed network, of its ancestors and descendants. We also assume that there is sufficient overlap between the cells in the same overlay. 3. 2 Performance Measures The objective of a load balancing protocol is to maximize the utilization of available resources [5]. In a wireless computing system, there are two very precious resources: the bandwidth of the wireless network and the battery power of the MHs. Therefore, the aim of our protocols is to maximize the bandwidth of the wireless network while minimizing the energy consumption of the MHs. This section ....

....for the same send and receive data rates over ffi t. Therefore, the total fraction of excess energy consumed by n MHs in s overlays is X total (ffit) k (ffit) 7) 3. 3 Load Balancing Protocols Our load balancing protocols can be classified as dynamic, distributed, cooperative algorithms [5]. By dynamic, we mean that the load balancing decisions are not made a priori, but vary over time, depending on the state of the network. Dynamic algorithms generally perform better than static algorithms, in realistic applications, because they do not rely on the fact that the resource ....

[Article contains additional citation context not shown here]

A. Goscinski, Distributed Operating Systems: The Logical Design. Sydney, Australia: Addison-Wesley Publishing Company, 1991.

....same order to all the participants on the network. Thus, causal ordering of the events is required in a multimedia collaborative system. The event ordering problem in distributed computing systems, including the logic clocks, the partial and total ordering of events, have been well studied [3][4]. However, the multimedia cooperative system has introduced new requirements concerning the causal ordering in the distributed system. The concurrency degree needed in a cooperative system varies in different applications. For instance, in joint editing, mutual exclusive accessing of the shared ....

A. Goscinski, "Distributed Operating Systems: The Logical Design", Addisor Publishing Company, 1991

....parallle gdndrique est composd de classes d objets au comportement similaire et dont les interactions entre objets reposent sur la nature des objets communicants et non pas sur leur identitd. Cette catdgorie de systbmes se retrouve aussi bien dans la recherche d algorithmes et de systbmes rdpartis [1] oh la gdndricitd est un facteur de toldrance aux pannes que dans la rdalisation de systbmes industriels oh elle est un facteur de simplification. Les techniques de description formelle (ESTELLE [2] LOTOS [3] se sont largement imposales comme mdthodes de spdcification car ddjh normalisdes ....

A. Goscinski, "Distributed Operating Systems. The Logical Design". Addison-Wesley 1991.

....because it provides the mapping between users and the SOLOs of an application. Access information Heirarchy of SOLOs, constructing a common inteoface Figure 3: Structure of an application ACCESS MODELS Access control is widely used within computing systems and many access models exist [5], many of which are derived from the Access Matrix Model [12] 6] In this general model (figure 4) there exists a set of passive Objects which are the resources to be protected, and a set of active Users (or Subjects) who wish to access these objects. Each User, Object pair (U, O) is associated ....

Goscinski , Andrzej "Distributed Operating Systems - The Logical Design" Addison-Wesley 1991 ISBN 0 201 41704 9. Chapter 11 Resource Protection. Pages: 585-647

....and programming paradigms. In particular, the assignment and schedule of the processors play a key role. Both of these problems are NP complete. Many approaches are discussed in literature addressing (dynamic) run time scheduling, static) compile time scheduling and the assignment problem [16, 14, 26]. Assuming that a program has already been written (or parallelized) using parallel programming primitives of a language or development environment, our approach consists in redistributing or migrating code from one processor to another one at run time, based on dynamic, adaptive and distributed ....

A. Goscinski. Distributed Operating Systems - Logical Design. Addison-Wessley, 1991.

....systems for parallel computers with distributed memory, for example Mach and Chorus already provide mechanisms for moving processes (or their corresponding constructions like actors etc. and thus for the integration of load management schemes (an overview of these systems can be found in [14,15]) The following section describes our approach of a load management system which uses migration of running processes as a mechanism for balancing the load. Thomas Ludwig Load Measurement Units Process Migration Units Load Evaluation Unit(s) 6 6 Hardware Operating System ....

A. Goscinski, Distributed Operating Systems -- The Logical Design (Addison-- Wesley, Sydney, 1991).

....currently existing dialogues with the object involved will be broken with a migration exception. Users performing invocations over these dialogues will thus detect that a nontransparent migration has occurred. 7. Related Work Many existing systems address the issues of object mobility [5, 8, 11]. A representative example is the Hermes system [3] which uses a variety of techniques to improve the efficiency of object invocation in the presence of location changes. A base technique is the use of forwarding addresses, which allow for transparent rerouting of invocation requests. Such ....

A. Goscinski, Distributed Operating Systems: The Logical Design, Addison-Wesley Publishing Company (1991).

....a single operating system. With distributed operating systems, users are not aware where their files are stored; nor are they aware that remote machines may execute their programs. All resources within the network are managed in a global fashion using global mechanisms rather than local mechanisms [1]. A group of machines could cooperate for a variety of reasons. A few of them are (1) Resource Sharing, 2) Performance Enhancement, 3) Reliability, 4) Fault Tolerance and (5) Transparency [1] Tens of distributed operating systems have been designed and implemented with various goals. Most ....

....within the network are managed in a global fashion using global mechanisms rather than local mechanisms [1] A group of machines could cooperate for a variety of reasons. A few of them are (1) Resource Sharing, 2) Performance Enhancement, 3) Reliability, 4) Fault Tolerance and (5) Transparency [1]. Tens of distributed operating systems have been designed and implemented with various goals. Most distributed system designs are willing to compromise on performance. On the other hand, systems that are designed to be performance oriented make no attempt to provide a single system image. They ....

A. Goscinski, "Distributed Operating Systems The Logical Design," Addison-Wesley Publishing Company, 1991.

....operating system. With distributed operating systems, users are not aware where their files are stored; nor are they aware that their programs may be executed by remote machines. All resources within the network are managed in a global fashion using global mechanisms rather than local mechanisms [1]. A group of machines could cooperate for a variety of reasons. A few of them are (1) Resource Sharing, 2) Performance Enhancement, 3) Reliability, 4) Fault Tolerance and (5) Transparency [1] Tens of distributed operating systems have been designed and implemented with various goals. Most ....

....within the network are managed in a global fashion using global mechanisms rather than local mechanisms [1] A group of machines could cooperate for a variety of reasons. A few of them are (1) Resource Sharing, 2) Performance Enhancement, 3) Reliability, 4) Fault Tolerance and (5) Transparency [1]. Tens of distributed operating systems have been designed and implemented with various goals. Most distributed system designs are willing to compromise on performance. On the other hand, systems that are designed to be performance oriented make no attempt to provide a single system image. They ....

A. Goscinski, "Distributed Operating Systems The Logical Design," Addison-Wesley Publishing Company, 1991.

....each with its own design goals and problem domain. In general, clustering systems tend to attack issues like performance, resource sharing, reliability, transparency, etc. Most of these design goals are conflicting and can never be accomplished in one system without some level of compromise [1]. Systems like Condor [2] Beowulf [3] provide highperformance environments while systems like Amoeba (distributed operating system) 4] tend to provide a good resource sharing environment. With these issues in mind, we are designing a distributed system, PODOS [5] an experimental Linux [6] ....

A. Goscinski, "Distributed Operating Systems The Logical Design," Addison-Wesley Publishing Company, 1991.

....Static distribution relies only on the hardware description and the prototype architecture. It computes a file which assign prototype components onto processors. Dynamic distribution takes also into account the current load of target processors during execution. Two approaches are proposed in [Goscinski 91] The task placement uses a theoretical load to sort logical processors in a configuration file. A mapping between logical processors and effective machines is deduced from the load at the beginning of the prototype execution. Software objects are placed as defined in the configuration file. ....

A. Goscinski, "Distributed Operating Systems : the logical design", chapter 8, Addison-Wesley, 1991.

....to utilise effectively the resources available. Many workstations that are either idle or lightly loaded are used to a greater extent in a distributed system, as the distributed operating system has mechanisms in place to balance the load of the system as a whole and produce a greater throughput [Gosc91]. Building a Parallelising Compiler for Distributed Computing Systems Chapter 2 Compilers and Parallelism 11 When two or more processes need to share information on a uniprocessor system they can easily do so through shared memory. Any race conditions that could occur between competing ....

....of semaphores, monitors, etc. Tane90] In a distributed system, however, when two or more processes need to share information, they must do so through the communication network. Other methods that are currently used for communication between processes are message passing or remote procedure calls [Gosc91]. Having to rely on message passing inherently depends on the underlying network and thus overhead cost of communication. Therefore parallel programs destined to execute on such a system must be of a large enough grain so that there isn t a loss in performance. Sharing resources between users in a ....

[Article contains additional citation context not shown here]

A.M. Goscinski. "Distributed Operating Systems - The Logical Design", AddisonWesley, 1991. 53

....sends only the absolute minimum state necessary to begin process execution on the destination host. The organization of kernel state required in framework design in Choices dramatically speeds up the excision and insertion of process state from the kernel. Results show that a system such as Accent [11] spends on average 17.5 of overall migration latency simply excising process state from different parts of the system [12] In contrast Choices spends only 3.6 of overall migration latency for the combined time to excise at the old 3 host and to insert at the new host. Since Accent s algorithm, ....

A. Goscinski. Distributed Operating Systems: The Logical Design. Addison-Wesley, Sydney, Australia, 1991.

....Static distribution relies only on the hardware description and the prototype architecture. It computes a file which assign prototype components onto processors. Dynamic distribution takes also into account the current load of target processors during execution. Two approaches are proposed in [Goscinski 91] Figure 3 : The prototype generic architecture. The Resource manager deals with Resources of the Petri net and must evaluate database like queries performed by clients (Processes) The Synchronization manager deals with some Actions of the model. The Resource and the Synchronization manager ....

A. Goscinski, "Distributed Operating Systems : the logical design", chapter 8, Addison-Wesley, 1991.

....logical objects 63 resent heterogeneous resources, and have to communicate with each of them frequently. This implies that the agent could be blocked for a large percentage of time. A multithreaded agent would resolve this problem to a substantial extent. By providing multiple threads of control [GOS91], communications can be dynamically delegated to threads, and the agent can continue internal processing tasks (such as event reporting and real object polling) or servicing requests from OSI managers. However, a multithreaded implementation needs additional considerations: n Since multiple ....

A. Goscinski, "Distributed Operating Systems: The Logical Design", Addison -Wesley Publishers Ltd., 1991.

....and multiprocessors. Potential reliability improvements are due to the data and control redundancy possible, the geographical distribution of the system, and the ability for hosts and communication processors to perform mutual inspection. With the proper subnet, distributed operating system [46], and distributed database [85] it is possible to share hardware and software resources in a cost effective manner, increasing productivity and lowering costs. Possibly the most important potential advantage of a DCS is extensibility. Extensibility is the ability to easily adapt to both short and ....

....large. For example, the Internet has over 500,000 nodes on it. This chapter has discussed two of the areas that played a major role in this achievement: distributed operating systems and distributed databases. For more information on distributed computing see the following good books and surveys [4, 35, 41, 46, 58, 85, 113, 115, 121]. As mentioned in the Introduction many areas of distributed computing could not be covered in this chapter. One important area omitted is distributed file servers such as NFS and Andrew. For more information on these and other distributed file servers see the survey article [68] 5 ....

A. Goscinski, Distributed Operating Systems: The Logical Design, Addison Wesley, Sdyney, 1991.

No context found.

Goscinski, A. (1991) Distributed Operating Systems: The Logical Design. Addison-Wesley Publishing Company, Reading, MA.

No context found.

A. Goscinski. "Distributed Operating Systems: The Logical Design". AddisonWesley Publishing. 1991.

....by many clients. For simplicity, we assume that each replica knows the location of other replicas that store the same data objects. This assumption can be loosened if a replication directory service is used. When requesting a service, a client specifies the service through, say,anattributed name [17]. The exact location of the service and the server that provides such a service will be determined by the system. We model a service as a set of replicas: S = fS 1 #S 2 # Delta Delta Delta#S i # Delta Delta Delta#S n g, where i =1# 2# Delta Delta Delta#n, are called the sequence numbers of ....

....data objects managed by the replica. WeusePto denote the set of all remote procedures provided byall replicas of the system: P = fpjp is a remote procedure of the systemg Our RPC model has the exactly once call semantics in the absence of failures and the at most once call semantics otherwise [17]. In particular, after wemakean RPC, the call may return successfully or fail. There maybeseveral reasons for the failure of a call, suchas object not free , server error , or communication error. With some error conditions it is clear to the client that the procedure was not executed, ....

A. Goscinski. Distributed Operating Systems: The Logical Design. Addison-Wesley Publishing Company, Reading, Massachusetts, 1991.

No context found.

Goscinski, A. (1991). Distributed operating systems: The logical design. Addison-Wesley: Sydney.

No context found.

: A. Goscinski, Distributed Operating Systems - The Logical Design, Addison-Wesley Publishing Company, 1992.

No context found.

A. Goscinski. "Distributed Operating Systems - The Logical Design", Addison-Wesley, 1991.

No context found.

Goscinski, A., Distributed Operating Systems: The Logical Design, AddisonWesley Publishing Company Inc. 1991.

No context found.

A. Goscinski. "Distributed Operating Systems - The Logical Design", AddisonWesley, 1991.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC