A. Barak, S. Guday, R. Wheeler, The MOSIX Distributed Operating System,Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672, SpringerVerlag, 1993  Home/Search   Document Not in Database   Summary   Related Articles   Check

....sequential jobs. All the above systems lack transparency as either special commands are introduced, or the user is require to rewrite, recompile and relink existing code with special libraries. Another approach that is totally transparent is to solve the problem at the kernel level. Both Mosix [1] and SSI for Linux [12] are such solutions. However, one needs to commit to the OS and there is a whole set of administrative considerations. Cluster Starter Kit for Linux [14] is an application for the management and monitoring of a Linux cluster. It allows administrators to monitor for certain ....

....communication between a process and a remote LM. The communication between any two LMs is peer to peer. This avoids the use of broadcast or any centralized server. 4.2 Algorithms used Dissemination of load information. The algorithm used in the dissemination of load information is given by Mosix [1], Algorithm 8.1. This involves updating the local load value and randomly sending to another node half of the local load vector entries. When the local node receives load information vector from another node, the former will merge the received information into the local load vector. Triggering ....

A. Barak, S. Guday, and R. Wheeler, The MOSIX Distributed Operating System, Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672, SpringerVerlag, 1993.

....pattern for each case. Keywords: Code Mobility, Mobile Agents, Wireless Mobile Computing, Mobility Management, Mobility Patterns. 1. INTRODUCTION In distributed systems, the notion of mobility evolved [1] from process migration (the act of transferring a process between two computers, e.g. [2], 3] 4] to mobile computing (the physical type of mobility which involves the movement of a physical device together with its owner, e.g. 5] 6] Mobility is inherent in wireless computing since users and their devices change location as they move. With the increasing demand of faster and more ....

Barak, A., Guday, S., and Weeler, R.G., The MOSIX Distributed Operating System, Load Balancing, UNIX, Springer Verlag, 1993.

....past without any major impact to distributed computing. Process migration has typically been used for load balancing in parallel processing systems and includes systems such as MPVM [5] and DynamicPVM [8] and Chime [27] Operating systems that support process migration include Chorus [22] MOSIX [3], Amoeba [27] and Sprite [11] However in these systems, the applications have restrictions, or have to be written for the particular OS, or have to be recompiled or re linked. Past attempts at process migration have been limited to processes without any I O connections or GUI handling. The lack ....

....and the proxy can be renegotiated when they migrate, but since the proxy does not migrate, the proxy to nonwrapped application connection remains the same. 5. Related Work There are several existing solutions to process migration problem on UNIX like MPVM [5] Condor [17] LibChkPt [20] MOSIX [3], and Sprite [11] To our knowledge, the systems that provide process migration on Windows NT are NT SwiFT [15] at Bell Labs and a checkpoint facility [24] developed at Intel, Israel. The following sections explain the related work. 5.1 MPVM Parallel Virtual Machine (PVM) is a software system ....

A. Barak, S. Guday and R. G. Wheeler. The MOSIX Distributed Operating System, Lecture Notes in Computer Science, Vol. 672, Springer, 1993.

....signals. The batches are handled out to contributor s machines that have downloaded and installed the 6 Seti Home client program. The client program runs as a screen saver on the contributor s machine and contacts the Seti servers for work when the machine has idle time. The Mosix system [BGW93, Mosix] offers a cluster based metacomputer. Each node in the cluster runs with a modified operating system. Processes are run on any node, and the system can dynamically move jobs around to other nodes in order to most efficiently use the resources of the system. Processes have no knowledge that they ....

Barak A., Guday S. and Wheeler R., The MOSIX Distributed Operating System, Load Balancing for UNIX, Springer-Verlag, 1993

....in Section 4.2) manage process groups and sessions. The process manager contains the node specific data. In particular, it holds a map from pids to virtual process objects for all local processes and processes originally created on this node. This is analogous to the home structure in MOSIX [3]. It also contains object references to the process managers on the other nodes in the cluster. Thus, the process manager is used to locate processes, to locate other nodes, and to perform operations that aren t associated with a particular process (e.g. sigsendset) To locate processes, we ....

.... from a node scheduled for maintenance; at this time we will provide rfork and migration (which are very similar, since a rfork is basically a fork and a migrate) The implementation of migration in Solaris MC will be similar to that in other distributed operating systems (e.g. Sprite [7] MOSIX [3], or Locus [20] moving the address space to the new node. The rexec call operates by packaging up the necessary state of the process (the list of open files, the proc t data, and the list of children) and sending this to the destination node along with the exec arguments. The process is then ....

[Article contains additional citation context not shown here]

A. Barak, S. Guday, and R. G. Wheeler, "The MOSIX Distributed Operating Systems," Lecture Notes in Computer Science 672, Springer-Verlag, Berlin, 1993.

....This formation allows user processes to run in a site independent manner. Processes in Mosix have the same structure as in UNIX, however, as Mosix supports migration, there are a few implications. Areas affected include: remote paging, locating other processes and interprocess communication [Barak et al. 93] The changes to these areas include: using a site independent reference to pages; extra fields and flags in the process table to allow load balancing and migration; and utilising a home node structure to find migrated processes. To migrate a process in Mosix: The source node allocates a ....

A. Barak, S. Guday, R. Wheeler. The MOSIX Distributed Operating System. Load Balancing for UNIX. Springer-Verlag.

....It might turn out that strict service neutrality is applicable only to few and uninteresting operations X. In similar situations, probabilistic methods have been very useful. For example, complexity theory uses probabilistic complexity; probabilistic properties are helpful in load distribution [1] and for probabilistic broadcasts [3] stochastic capacities [4] describe cache properties independent of specific benchmarks. Accordingly, we define probabilistic service neutrality as a weaker but more powerful concept: ffl Operations X(x j ) are service P neutral iff there exists an X free ....

A. Barak, S. Guday, and R. Wheeler. The MOSIX Distributed Operating System, Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672. Springer--Verlag, 1993.

....to either the application programs or the underlying op erating system. For instance, GLUnix [Gh 98] is a set of programs and library functions to provide applications running on a network of Unix workstations with a uniform view of distributed resources. In other projects, such as MOSIX [Ba 93] and SPIN [Be 95] operating systems have been modified to support distributed computations and to provide transparent process mobility. More recent approaches have relied on extending standard operating systems, intercepting operating system calls made by applications, and manipulating these ....

Amnon Barak, Shai Guday and Richard G. Wheeler, The MOSIX Distributed Operating System, Lecture Notes in Computer Science, Vol. 672, Springer, 1993.

....indeed see it as a single shared memory machine rather than as a set of machines having their own resources. Although memory management (remote paging [8, 7] SVM [16, 1, 20] disk management (RAID [5] parallel or distributed file system [6, 21, 23] processor management (process migration [24, 22]) have already been extensively studied, very few work has been done for global resource management in which these mechanisms cooperate or are integrated towards the objective of executing high performance applications. Moreover, we are convinced that high availability is a key property to be ....

A. Barak S. Guday R. Wheeler. The MOSIX Distributed Operating System, volume 672 of LNCS. Springer Verlag, 1993.

....control and resource allocation. In a context where every job has a reward associated with it, and every producer has a cost, using either virtual or real money, these strategies will provide competitive profits and a measure of security against system failures. Related Work The Mosix system [BGW93] provides a set of enhancements to Unix which incorporate the CPU resources of a Network of Workstations by enabling preemptive migration of processes and dynamic load balancing within a network that has installed Mosix. These enhancements are completely transparent to the application level. Mosix ....

....Monitor applets. Experimental Results The first major test of the Java Market was performed in the Johns Hopkins Center for Networking and Distributed Systems (CNDS) A CPU intensive simulation (about 1000 lines of Java code) evaluating five different job scheduling policies for the Mosix system [BGW93] and a stream of jobs was run on the CNDS lab machines. One hundred simulations, each of them representing 10,000 real time seconds, were run in the following two ways: 1. Running on a standalone Pentium II machine using the Java Developer s Kit. 11 2. Using the Java Market with six producer ....

A. Barak, S. Guday and R. Wheeler. The Mosix distributed operating system, load balancing for Unix, Volume 672, May 1993.

....Monitor applets. Experimental Results The first major test of the Java Market was performed in the Johns Hopkins Center for Networking and Distributed Systems (CNDS) A CPU intensive simulation (about 1000 lines of Java code) evaluating five different job scheduling policies for the Mosix system [7] and a stream of jobs was run on the CNDS lab machines. One hundred simulations, each of them representing 10,000 real time seconds, were run in the following two ways: 1. Running on a standalone Pentium II machine using the Java Developer s Kit. 2. Using the Java Market with six producer ....

....for networks of workstations where jobs can only be assigned to a machine once, it distributes jobs using a straightforward round robin policy. Programmers can override this policy. In comparison, our strategy for an identical network of workstations completed the average job 38 faster. Mosix [7] is a set of kernel enhancements to the BSDI Unixlike operating system [9] that allows jobs to be moved from machine to machine without interrupting their execution. Mosix also has an experimentally tuned resource allocation strategy based on load balancing. A Cost Benefit based strategy for Mosix ....

A. Barak, S. Guday and R. Wheeler. The Mosix distributed operating system, load balancing for Unix, Volume 672, May 1993.

....bus stops, but important modifications of the programs compilation process seem to be required. Although numerous studies related to load balancing have been conducted in the last decade, few practical systems have been built. The most popular systems are Utopia [25] Condor [15] Mosix [1], and Calypso [2] Utopia [25] built at the university of Toronto, is a system for automatic load sharing in large, heterogeneous distributed systems. Utopia, unlike Stardust, does not support process migration; it only supports initial process placement. The Condor system, at the university of ....

....provides load sharing of sequential batch jobs [15] Condor aims at making use of idle workstations. In contrast, Stardust supports both batch and interactive jobs and makes use of not only idle workstations. Like Stardust, Condor supports process migration, but only on homogeneous hosts. Mosix [1] is a distributed operating system built at Hebrew University of Jerusalem. It is an enhancement of Unix with network transparency and automatic dynamic load balancing. Since load balancing is implemented inside the Mosix kernel, there is no problem while migrating OS dependent data structures. ....

Barak, A., Guday, S., and Wheeler, R. The MOSIX Distributed Operating System, Load Balancing For Unix. No. 672 in Lecture Notes in Computer Science. Springer Verlag, 1993.

....our experience with it in another setting a classic network of workstations. In this context, Enhanced Mosix (and its cousin Enhanced PVM) are strategies for job allocation and reallocation in a scalable computing cluster. We have developed these strategies in a joint work with the Mosix group [1, 2, 3, 4]. Each of these test cases has been shown to produce very good performance in practice. The Java Market, using machines connected only by the Internet, produced a dramatic increase in our sample application s speed. Because the machines used only Web browsers to connect to the Market, they could ....

....for resources. This approach uses economic principles conceptually similar to ours. The Cost Benefit framework, however, integrates the computer science notions of competitive algorithms and analysis with the economics concepts of marginal costs and markets for services. The Mosix system [2, 4] enhances several operating systems with the ability to transparently move jobs from one machine to another. This allows the system to quickly correct all serious mistakes in job allocation. Mosix is a powerful system independently and can be combined with our resource allocation strategies for ....

[Article contains additional citation context not shown here]

A. Barak, S. Guday and R. Wheeler. The Mosix distributed operating system, load balancing for Unix, Volume 672, May 1993.

....prior to regular shutdown, or through checkpointing [14, 18] A variety of process and task migration systems have been developed. All these systems have limitations that make them usable only for small classes of applications or in special environments. Most run on experimental operating systems [2, 3, 5, 20], on special architectures [15] or need modifications to standard operating system kernels [1, 18, 27] Migration based on a special operating system is difficult to apply to existing workstation clusters, especially in production environments. Migration systems that are implemented in user ....

....running the NORMA version of Mach across an Ethernet are given as 0.5 seconds resp. 0.25 seconds by Milojici c [18] Also, migration times for the distributed operating systems Sprite, V and Charlotte are cited in [18] to be in the same order of magnitude. For the MOSIX system, Barak and Wheeler [3] report transfer rates of about 600 800KB s for process migration over a ProNET 80 token ring between machines with NS32532 CPUs Published in Operating Systems Review, Vol. 29, No. 4, October 1995, pg. 0 50 100 150 200 250 0 1000 2000 3000 4000 5000 6000 7000 8000 Total process size [KB] NFS ....

[Article contains additional citation context not shown here]

Amnon Barak, Shai Guday, and Rich ard G. Wheeler. The MOSIX Distributed Operating System, volume 672 of LNCS. Springer, 1993.

....Due to the problems cited above, we initiated a project to investigate alternative cluster technologies. Given one author s past experience with an earlier version of FDDI Ring 26 SUN Ultra 2s SCSI 2 EMC Celerra NFS Servers 2 Symmetrix Systems Figure 1: The Original Cluster MOSIX [Barak et al. 1993], we tried to use the Linuxbased MOSIX cluster at Hebrew University [MOSIX] to build our code images. MOSIX adds load information about the cluster, process migration, and various other clustering features to Linux without adding any new, clusterspecific APIs. This allows us to continue to use ....

Barak, A., Guday, S., and Wheeler, R. (1993) The MOSIX Distributed Operating System. Lecture Notes in Computer Science, Vol. 672, SpringerVerlag.

.... executions, but perform process migration at the explicit request of the user [4] Load balancing algorithms for SCC s were extensively researched in recent years and many schemes were simulated [5, 6] In spite of the diversity of suggested methods, only few systems implement load balancing [7, 8] or load sharing [4] by preemptive process migration, and even fewer systems address other resource sharing in a NOW [1] The availability of a preemptive process migration mechanism for dynamic work distribution creates an opportunity to manage other resources in an SCC. This paper presents a ....

....could be based on the MS algorithm, in which the migrated process is placed in a node chosen from an information window. This is a cache that holds information on the free memory of other nodes. The window could be updated at regular intervals, using a suitable information dissemination algorithm [7]. The relevant parameters of this algorithm are the size of the window and the rate of the information dissemination. Simulations of the above scheme, called Windowed MS are shown in Figure 4. 80 82 84 86 88 90 92 94 96 98 100 4 8 16 32 64 128 256 512 Number of Nodes 32 16 8 80 82 84 86 88 90 ....

[Article contains additional citation context not shown here]

A. Barak, S. Guday, and R.G. Wheeler. The MOSIX Distributed Operating System, Load Balancing for UNIX. In Lecture Notes in Computer Science, Vol. 672. Springer-Verlag, 1993.

....which combined with load balancing can maximize the overall performance, respond to resource availability and achieve high degree of overall utilization of the NOW resources. In spite of the advantages of process migration and load balancing, there are only few systems that support these services [2, 7, 9]. The main reason is the fact that most parallel programming environments are implemented above the operating systems and are geared to support heterogeneous configurations. For example, p4 [5] is a library of macros and routines for programming a wide range of parallel machines, including ....

....a per machine basis. Dynamic process creation is limited to process spawning in the local host by a preassigned parent process. This paper presents the performance of executing sev eral benchmarks using PVM, with its static process assignment vs. PVM with the MOSIX preemptive process migration [2]. PVM [8] is a popular programming environment which lets users exploit collections of networked computers and parallel computers. Its main advantages are the support of heterogeneous networks and machines, dynamic process and virtual machine management, and a simple and efficient user interface ....

[Article contains additional citation context not shown here]

A. Barak, S. Guday, and R. Wheeler. The MOSIX Distributed Operating System, Load Balancing for UNIX. In Lecture Notes in Computer Science, Vol. 672. SpringerVerlag, 1993.

....fix such a mistake. It is intuitively clear that the ability to migrate jobs could lead to better performance that is, faster completion times for the average job. Unless it is known where a job should be at any given time, however, the reassignment strategy could also make mistakes. The Mosix [BGW93, BL97] kernel enhancements to the BSD OS Unix like operating system [Bsdi] for example, allow this kind of transparent job migration. Determining the optimal location for a job is a complicated problem. The most important complication is that the resources available on a cluster of workstations are ....

A. Barak, S. Guday and R. Wheeler. The Mosix distributed operating system, load balancing for Unix, Volume 672, May 1993. 13

....of MOSIX and its main characteristics. Section 3 presents the performance of its communication protocol and the resource sharing algorithms. Section 4 presents the performance of several, large scale parallel applications. Our conclusions are given in Section 5. 2 Overview of MOSIX MOSIX [4] is a set of enhancements of BSD OS [5] with algorithms for adaptive resource sharing. These algorithms are geared for efficient resource utilization among the (homogeneous) nodes of a distributed memory, shared nothing scalable CC, including LAN connected networks of workstations and servers. ....

A. Barak, S. Guday and R.G. Wheeler, The MOSIX distributed operating system, load balancing for UNIX, Lecture Notes in Computer Science 672 (Springer-Verlag, Berlin, 1993).

No context found.

A. Barak, S. Guday, R. Wheeler, The MOSIX Distributed Operating System,Load Balancing for UNIX. Lecture Notes in Computer Science, Vol. 672, SpringerVerlag, 1993

No context found.

Barak, A., Guday, S. and Wheeler, R., 1993. The MOSIX Distributed Operating System, Load Balancing for UNIX.

No context found.

A. Barak, S. Guday, and R. G. Wheeler. The MOSIX distributed operating system. 672, 1993.

No context found.

A. Barak S. Guday R. Wheeler. The MOSIX Distributed Operating System, volume 672 of LNCS. Springer Verlag, 1993.

No context found.

A. Barak, S. Guday, and R. G. Wheeler. The Mosix Distributed Operating System. Lecture Notes in Computer Science 672, Springer-Verlag, Berlin, 1993.

No context found.

A. Barak, S. Guday, and R. G. Wheeler. The Mosix Distributed Operating System. Lecture Notes in Computer Science 672, Springer-Verlag, Berlin, 1993.

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