Barak A, La'adan O, Shiloh A. Scalable cluster computing with MOSIX for LINUX. Proceedings of the 5th Annual Linux Expo, May 1999; 95--10.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....nodes and runs entirely in kernel space on an unmodified Linux kernel; applications can take advantage of network swapping without having to re compile or link with special libraries. Nswap is designed to scale to large clusters using an approach similar to Mosix s design for scalability [1]. In Nswap there is no centralized server that chooses a remote node to which to swap. In addition, Nswap does not rely on global state, nor does it rely on complete or completely accurate information about the state of all cluster nodes to make swapping decisions. Thus, Nswap will scale to larger ....

Barak A., La'adan O., and Shiloh A. Scalable cluster computing with MOSIX for Linux. In Proceedings of Linux Expo '99, pages 95--100, Raleigh, N.C., May 1999.

....will be much faster in a large cluster. 5 Related Works There are many ongoing projects related to the cluster middleware. Basically, these projects can be classified into two approaches. The first approach is to extend operating system functionality at kernel level. Some example works are Mosix[5], Nomad [6] and Bproc [7] Mosix offers a kernel level, adaptive resource sharing algorithms and low overhead process migration. Nomad effort is quite similar to Mosix but has a wider scope of building a scalable cluster OS that provides a singlesystem image, resources management, and distributed ....

A. Barak, O. La'adan, and A. Shiloh, "Scalable Cluster Computing with MOSIX for LINUX", in Linux Expo 99, pp95-100, Raleigh, N.C., USA, May1999

....systems, the AppLeS project [8] provides applicationlevel scheduling agents on heterogeneous systems, taking into account their actual resource performance. Another approach is to include cluster computing capabilities into the operating system itself. MOSIX (Multicomputer OS for UNIX) [4] is such a software package. It enhances the Linux kernel and allows any size cluster of X86 Pentium 20 based workstations and servers to work cooperatively as if part of a single system. Its main features are preemptive process migration and supervising algorithms for load balancing and memory ....

A. Barak, O. La'adan, and A. Shiloh. Scalable Cluster Computing with MOSIX for LINUX. Proc. Linux Expo '99, Raleigh, N.C., May 1999, pp. 95-100.

....can be used to gain a better understanding of the operating system internals and behavior. But even more important is the option of modifying the operating system. For example, the Mosix system provides a load balancing capability among the nodes of a cluster, based on a process migration facility [7]. This is implemented within the kernel of the base system, which is Linux. Likewise, the development of Beowulf hinged on kernel modifications that allowed several physical networks to be used as a single logical network [36] In Linux, all the source code is available under the terms of the GNU ....

A. Barak, O. La'adan, and A. Shiloh, "Scalable cluster computing with MOSIX for Linux ". In Linux Expo, pp. 95--100, May 1999.

....is created. The shadow inherits all resources used by the parent. After the migration, processes use resources through the shadow. This solution is transparent but rather expensive. The important disadvantage of the shadow is that it is using the host where the process was initiated. Mosix [1] is a software to support the cluster computing and it is implemented on the operating system level. Each Unix version requires di erent implementation of Mosix, and recent 7 th implementation of Mosix was developed for Linux using the x86 based processors. The Mosix migration mechanism is ....

....indirect routing mode [4, 6] after migration the process leaves a piece of code on the old machine and this code takes care of forwarding data to the new process location. In this paper term mirror is used for a process which is used to redirect network packets. This concept was suggested in [1, 8]. Neither solution is perfect. Global (centralized) data distribution is not fault tolerant and could be too slow. In the second case, after process migration the machine is still in use, but in a di erent way. It means that the machine can not go down. This solution is not fault tolerant, too. ....

A. Barak, O. La'adan, and A. Shiloh. Scalable Cluster Computing with MOSIX for LINUX. In Proceedings of Linux Expo 1999, pages 95100, May 1999. http://www.mosix.org/.

....GLUnix, Butler, HetNOS, and LSF, for example, allow users to write applications which link with a C library of remote execution related functions. Using this model, applications such as a shell which automatically decides whether to execute a job locally or remotely have been developed. MOSIX [2], NEST [1] COCANET Unix [16] and Solaris MC [10, 17] are examples of modifying and extending an existing operating system. Again, each of the systems supports only a subset of REXEC s features. NEST, for example, supports transparent remote execution but does not support features such as dynamic ....

BARAK, A., LA'ADAN, O., AND SMITH, A. Scalable cluster computing with mosix for linux. In Proceedings of Linux Expo '99 (May 1999), pp. 95--100.

No context found.

Barak, A., La'adan, O. and Shiloh, A., " Scalable cluster computing with MOSIX for Linux ", Proc. 5-th Annual Linux Expo, Raleigh, NC, pp. 95-100, May 1999.

....Method The execution cluster consists of 6 machines, two Pentium Pro 200MHz; two Pentium II 300MHz and two Pentium II 400MHz, with memory sizes ranging from 64MB to 256MB. The machines where connected by a Fast Ethernet switch. The executions were performed using a modified MOSIX for Linux [8] operating system, which supports preemptive process migration. In each execution, we replaced the existing resource sharing policy of MOSIX with the COST P or the GREEDY P policies respectively. 11 The implementation of the COST P algorithm is a straight forward implementation of the COST IO ....

....we use a limited number of task reassignments, i.e. moving a task from one machine to another after it has started to execute. We note that experience with real systems shows that task reassignments may be performed with very little overhead, thus achieving a powerfull tool for load balancing [7, 8]. Task reassignments are also used for temporary tasks on unrelated machines [4] 4.2 Unit Costs at Home In this section we consider a variant of the home model in which the cost vector is of the following form: 8i; 1 i m : p j (i) 1 i = h j p j otherwise ; 18 AssignH(j, p j , l, ....

[Article contains additional citation context not shown here]

A. Barak, O. La'adan, and A. Shiloh. Scalable cluster computing with mosix for linux. In Proc. 5-th Annual Linux Expo, pages 95--100, May 1999. http://www.mosix.cs.huji.ac.il.

....competitive ratio, we use a limited number of job reassignments, i.e. changing the assignment of a job from one machine to another. We note that experience with real systems shows that job reassignments may be performed with very small overhead, thus achieving a powerful tool for load balancing [5, 6]. Job reassignments are also used in the (theoretical) algorithm for temporary jobs on unrelated machines [3] 2.2 Unit Loads at Home We consider a variant of the home model in which the load vector s form is: 8i; 1 i m : p j (i) 1 i = h j p j otherwise ; such that 8j : p j 1. ....

....Policy 9 summarizes the results, showing the performance of the heuristics that are implemented for the rest of this section. We note that the original policy performed an average of one reassignment for every nine jobs. Results from real clusters, e.g. the MOSIX distributed operating system [5, 6], indicate that process migration is not very expensive, thus implying that such tradeo of performance improvement and limited number of reassignments is worth while. We also note that all the load balancing algorithms in our simulations are allowed to make reassignments (e.g. greedy job ....

A. Barak, O. La'adan, and A. Shiloh. Scalable cluster computing with mosix for linux. In Proc. of the 5th Annual Linux Expo, pages 95-100, May 1999. http://www.mosix.cs.huji.ac.il.

....of MFS with and without DFSA vs. NFS. Our conclusions are given in Section 5. 2 2 MOSIX Background MOSIX is a software package that enhances the Linux kernel with cluster computing capabilities. Its main features are preemptive process migration and supervising algorithms for load balancing [4] and memory ushering [3] MOSIX operates silently and its operations are transparent to the applications, just like in an SMP. Users can run parallel (and sequential) applications by initiating processes in one node then allow the system to assign and reassign processes to the best available nodes ....

A. Barak, O. La'adan, and A. Shiloh. Scalable Cluster Computing with MOSIX for LINUX. In Proc. 5-th Annual Linux Expo, pages 95--100, May 1999.

No context found.

Barak A, La'adan O, Shiloh A. Scalable cluster computing with MOSIX for LINUX. Proceedings of the 5th Annual Linux Expo, May 1999; 95--10.

No context found.

A. Barak, O. La'adan, A. Shiloh. Scalable cluster computing with MOSIX for Linux. In Proceedings of the 5th Annual Linux Expo, p 95--100, 1999.

No context found.

A. Barak, O. La'adan, A. Shiloh, "Scalable Cluster Computing with MOSIX for LINUX", Proc. Linux Expo '99, Raleigh, N.C., May 1999, pp. 95-100.

No context found.

A. Barak, O. La'adan, A. Shiloh. Scalable cluster computing with MOSIX for Linux. In Proceedings of the 5th Annual Linux Expo, p 95--100, 1999.

No context found.

BARAK, A., LA'ADAN,O.,AND SHILOH,A. 1999. Scalable cluster computing with MOSIX for Linux. In Proceedings of the 5th Annual Linux Expo. Raleigh, N.C. (May), 95--100.

No context found.

O. t. Amnon BARAK and A. SHILOH. Scalable cluster computing with MOSIX for LINUX. In Proc. Linux Expo '99, pages 95--100, May 1999.

No context found.

BARAK, A., LA'ADAN, O., AND SMITH, A. Scalable cluster computing with mosix for linux. In Proceedings of Linux Expo '99 (May 1999), pp. 95--100.

No context found.

A. Barak, O. La'adan, and A. Shiloh. Scalable cluster computing with mosix for linux. In Proceedings of the 5-th Annual Linux Expo, pages 95100, May 1999.

No context found.

Barak A, La'adan O, Shiloh A. Scalable cluster computing with MOSIX for Linux. Linux Expo, May 1999; 95--100.

No context found.

La'adan. O Barak. A and Shiloh. A., "Scalable Cluster Computing With MOSIX for Linux," Journal of Future Generation Computer Systems, pp. 361--372, May 1998.

No context found.

A. Barak, O. La'adan, and A. Shiloh. Scalable Cluster Computing with MOSIX for LINUX. In Linux Expo '99, 1999.

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