Douglas Comer and James Grioen. \A New Design for Distributed Systems: The Remote Memory Model". In Proceedings of the Summer 1990 USENIX Conference, pages 127-135. USENIX Association, June 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to understand the features and limitations of yet another version of the application. The use of remote memory as a store for client device overconsumption has been investigated in many different forms over the last decade. They focused around the use of virtual memory paging to a remote server [10], 26] 11] to name but a few. Given their use of virtual memory mechanisms, these works assume a generic view of the memory usage semantics by applications using the system. In addition, much work has been done on distributed shared memory [19] and process object migration [22] which address ....

J. G. Douglas Comer. A new design for distributed systems: The remote memory model. In Proc. Summer 1990.

....system to effectively schedule page requests across different active processes. 2 Related Work The idea of remote virtual memory, particularly distributed and or shared, has been an ongoing subject for over 15 years. Comer and Griffioen examine the usefulness of a dedicated memory server in [7]. They make the distinction of separating the paging operation from the file backing store operation. Then they can focus on designing efficient memory and file servers. Another approach views the sum total memory of a cluster as a single cache space [10] Dahlin et al. suggest utilizing the ....

D. Comer and J. Griffioen. A new design for distributed systems: The remote memory model. In Proc. Summer

....the contents of traditional software RamDisks are lost in the event 20 of a crash. Several research groups have proposed various methods of exploiting network main memory to avoid disk accesses. In remote memory paging systems, for example, applications use the remote memory as swap area [3, 4, 8, 12, 16, 25]. As a generalization to remote memory paging, several researchers propose to use the remote main memory as a file system cache [2, 9, 10, 11, 15] For example, Feeley et al. have implemented a global memory management system (GMS) in a workstation cluster, using the idle memory in the cluster to ....

D. Comer and J. Griffioen. A new design for Distributed Systems: the Remote Memory Model. In Proceedings of the USENIX Summer Conference, pages 127--135, 1990.

....are very expensive, the memory server system overhead and the average page miss penalty with MS 0 are much greater than with MS 1. Consequently, MS 1 runs twice as fast as VM. 6 Related Work Researchers in the local area network community haven taken a similar approach to the memory server model [4, 9, 21, 8], using the memory available on other nodes as backing store. All these studies use a cluster of workstations and run sequential or distributed applications, while our memory server model runs on a multicomputer for sequential and parallel applications. Therefore, we have di#erent design ....

D. Comer and J. Gri#oen. A New Design for Distributed Systems: The Remote Memory Model. In Proceedings of the USENIX Summer Conference, pages 127--135, Anaheim, California, June 1990.

....system to effectively schedule page requests across different active processes. 2 Related Work The idea of remote virtual memory, particularly distributed and or shared, has been an ongoing subject for over 15 years. Comer and Griffioen examine the usefulness of a dedicated memory server in [7]. They make the distinction of separating the paging operation from the file backing store operation. Then they can focus on designing efficient memory and file servers. Another approach views the sum total memory of a cluster as a single cache space [10] Dahlin et al. suggest utilizing the ....

D. Comer and J. Griffioen. A new design for distributed systems: The remote memory model. In Proc. Summer

....and virtual memory swapping can gain substantially from using remote memory instead of disk. This has been used in several systems [13] With the remote memory server model, the large memories of dedicated nodes are used for swap space to lower the load of workstation with heavy paging activity [17]. With the N Chance Forwarding algorithm [18] remote memory improves distributed file system performance by a coordination of the contents of client caches. When the nodes are running low on physical memory, and after all replica blocks are removed, the locally oldest block is forwarded to a ....

D. Comer and J. Griffioen. A new design for distributed systems : The remote memory model. Proceedings of the 1990 USENIX Conference, June 1990.

....DIMMs cost 600 [21] Second, commodity priced machines have a limit on the amount of memory that can be installed; server class machines that can have larger memories come at a substantial premium. Several research projects have proposed using memory of idle workstations for hosting guest data [7, 8, 11, 13, 20, 22, 27]. Iftode et al. [13] propose extending the memory hierarchy of multicomputers by introducing a remote memory server layer; Schilit and Duchamp [23] investigated the use of remote memory paging for diskless portable machines; Dahlin et al. [8] and Sarkar et al. [22] propose schemes to use idle memory ....

D. Comer and J. Griffioen. A new design for distributed systems: The remote memory model. In Proceedings of the 1990 USENIX Summer Conference, 1990.

....memory system developed at the University of Washington. It is a Unix kernel facility that provides coordinated memory management for a cluster of autonomous computers, by managing the memories of cluster 2 nodes as a shared, distributed page cache. The GMS implementation supports remote paging [5, 10] and cooperative caching [7] of file blocks and virtual memory pages, unified at a low level of the operating system kernel. The purpose of GMS is not to support a shared memory abstraction, but rather to transparently improve the performance of data intensive workloads. The key insight is that ....

D. Comer and J. Griffioen. A new design for distributed systems: the remote memory model. In Proceedings of the 1990 Summer USENIX, June 1990.

.... with multiple architectures [19, 27] support ecient checkpointing [17] support limited forms of parallelism [2, 4, 6, 7, 24, 25, 26] and minimize network demand [5] Furthermore, opportunistic scheduling ideas have also been applied to harvesting idle remote memory in distributed memory systems [3, 9, 10, 12, 15, 23, 28, 29]. The primary goal of this past work has been to develop mechanisms to maximize throughput and add user functionality. Little work has been done on how best to schedule batch jobs. As shown from time sharing systems, the scheduling algorithm can have a signi cant impact on user satisfaction. One ....

D. Comer and J. Grioen. A new design for distributed systems: The remote memory model. In Proc. of the 1990 USENIX Summer Conference, 1990.

....used by these systems [17, 3] As mentioned in previous sections, the memory server model can be incorporated into an SVM system to exploit the utilization of available physical memory. Researchers in the local area network community haven taken a similar approach to the memory server model [8, 13, 30], using the memory available on idle workstations as backing stores. Although conceptually their approach is similar to that in this paper, there are several differences. Mainly, the memory server mechanism for multicomputers presented in this paper supports multiple programming models: ....

Douglas Comer and James Griffioen. A New Design for Distributed Systems: The Remote Memory Model. In Proceedings of the 1990 USENIX Summer Conference, pages 127--135, June 1990.

....use of memory on remote hosts. Some have seen it as a way for improving local performance via extending the amount of physical memory available to processes, while others have considered it the solution to the disk latency problem. We shall now look at some of these approaches. Comer and Grioen [6] describe a remote memory model that uses dedicated memory servers that clients could allocate memory on. The remote memory servers could extend their capacity by utilising a local disk. This would be transparent to clients who would not know where their data was coming from. The remote memory ....

....to solve, or even reduce, the e ects of the disk latency problem. However, the authors did suggest a machineindependent communication protocol that could be utilised by a remote paging system. Indeed the protocol used for the transfer of pages in my system makes use of the ideas described in [6]. The Dodo system [15] provides a method of using remote memory as an intermediate cache between memory and the local disk. 6 The Dodo system is provided via application level libraries and does not change any of the OS code. This makes the system more portable but also limits the ....

[Article contains additional citation context not shown here]

Comer, D., and Griffioen, J. A New Design for Distributed Systems: The Remote Memory Model. In Proceedings of the USENIX 1990 annual technical conference: June 11-15, 1990, Anaheim, California, USA (June 1990), USENIX, USENIX, pp. 127-135.

....Memory Servers) are dedicated to receive memory requests from other workstations (clients) which have not enough physical memory available. Clients may send their requests to several servers. If the entire main memory of a server is occupied, a local disk can be used. This method is presented in [6] and a simplified version is implemented. The second approach (Uniform Vision of the Servers) assumes that a workstation may be a server or a client of memory depending on the amount of free physical memory locally available. The advantages of this approach compared to the previous one are ....

Comer, D., Griffioen, J.: A New Design for Distributed Systems: The Remote Memory Model. In Procceedings of the USENIX Summer Conference. Anaheim, California. (June 1990) 127--135

....change, in turn, should enable us to build a new generation of 2 novel, high performance distributed applications. Consider the following two motivating examples. In a cluster of networked workstations, paging to spare memory lying idle on other nodes could be much faster than paging to disk [20]. A distributed server could manage the idle memory in the network; on a page fault, the page fault handler would communicate with the server and transfer pages to or from remote nodes. The effectiveness of this scheme is highly influenced by the communication performance between client and ....

Douglas Comer and James Griffioen. A new design for distributed systems: The remote memory model. In Proceedings of the Summer 1990 USENIX Conference, pages 127--135, June 1990.

....accepts paging requests from remote nodes. This system allowed local users to statically restrict the amount of physical memory available to the paging server. Comer and Griffioen described a remote memory model in which the cluster contains workstations, disk servers, and remote memory servers [19]. The remote memory servers were dedicated machines whose large primary memories could be allocated by workstations with heavy paging activity. No client to client resource sharing occurred,except through the servers. Felten and Zahorjan generalized this idea to use memory on idle client machines ....

Douglas Comer and James Griffioen. A new design for distributed systems: The remote memory model. In Proceedings of the USENIX Summer Conference, June 1990.

....systems. 2.1. Importance of Latency for NetworkMemory Our first use of Trapeze is as a messaging substrate for the Global Memory Service (GMS) 9] a Unix kernel facility that manages the memories of cluster nodes as a shared, distributed page cache. The GMS implementation supports remote paging [5, 10] and cooperative caching [6] of file blocks and virtual memory pages, unified at a low level of the operating system kernel. The purpose of GMS is not to support a shared memory abstraction, but rather to transparently improve the performance of data intensive workloads. GMS coordinates memory ....

D. Comer and J. Griffioen. A new design for distributed systems: the remote memory model. In Proceedings of the 1990 Summer USENIX, June 1990.

.... 1992] transparently compressing file systems [Burrows et al. 1992, Douglis 1993] load sharing [Zhou et al. 1992] process migration [Theimer et al. 1985, Douglis Ousterhout 1991] fast communication primitives [Bershad et al. 1990, von Eicken et al. 1992] upcalls [Clark 1985] network paging [Comer Griffioen 1990], and user level pagers [Young et al. 1987] There are two primary reasons for this. First, the complexity of modern operating system kernels often makes novel extensions very difficult to implement and maintain [Douglis Ousterhout 1991] Second, significant internal differences between kernels ....

D. Comer and J. Griffioen. "A New Design for Distributed Systems: The Remote Memory Model". In Proceedings of the 1990 USENIX Summer Conference, pp. 127-- 135, June 1990.

....of client applications starts to approach the aggregate size of the client memories. 6 Related Work Cooperative caching for file systems developed from research involving remote memory usage. The idea of remote memory servers in distributed systems was first introduced by Comer and Griffioen in [Comer90]. Felten and Zahorjan proposed the use of idle machines as remote memory servers[Felten91] Franklin et al. in [Franklin92] introduced the concept of remote client servers to extend the traditional client server database architecture. Leff et al. in [Leff91] showed that memory must be dynamically ....

Douglas E. Comer and J. Griffioen. A New Design for Distributed Systems: The Remote Memory Model. In Proceedings of the Summer 1990 Usenix Conference, pages 127--135, June 1990.

....low latency with high bandwidth. Since the Myrinet firmware is customer loadable, any Myrinet network site with PCI based machines can use Trapeze. GMS [14] is a Unix kernel facility that manages the memories of cluster nodes as a shared, distributed page cache. GMS supports remote paging [8, 15] and cooperative caching [10] of file blocks and virtual memory pages, unified at a low level of the Digital Unix 4.0 kernel (a FreeBSD port is in progress) The purpose of GMS is to exploit high speed networks to improve performance of data intensive workloads by replacing disk activity with ....

D. Comer and J. Griffioen. A new design for distributed systems: the remote memory model. In Proceedings of the 1990 Summer USENIX, June 1990.

....comes the problem of maintaining consistency. Addressing this problem has proven to require overhead inefficiencies and a good deal of complex code for a relatively uncommon situation. Also, distributed VM systems usually require a static partitioning of the shared address space. One previous work [2] suggests the use of remote memory as a new model of computation. The system described therein exploits a predeclared set of dedicated remote memory servers, and concentrates on the network protocols necessary for interoperability. We get a certain (lesser) degree of interoperability from Mach ....

Douglas Comer and James Griffioen. A new design for distributed systems: the remote memory model. In Proceedings, Usenix Summer Conference, pages 127--135, Anaheim, California, June 1990.

No context found.

Douglas Comer and James Grioen. \A New Design for Distributed Systems: The Remote Memory Model". In Proceedings of the Summer 1990 USENIX Conference, pages 127-135. USENIX Association, June 1990.

....of a network of workstations is often idle. Off the shelf systems provide access to this idle remote memory an order of magnitude faster than disk latencies. Therefore, many systems have been developed to make use of idle memory capacity, primarily for paging and caching. Comer and Griffioen [7] introduced the remote memory model in which client machines that exhaust their local memory capacity paged to one of a set of dedicated remote memory servers. Each client s memory was private and inaccessible even if it was idle. Data migration between servers was not supported. Felten and ....

....even if it was idle. Data migration between servers was not supported. Felten and Zahorjan [11] enhanced the remote memory model to use any idle machine. Idle client machines advertise their available memory to a centralized registry. Active clients randomly chose one idle machine to page to. Like [7], data was not migrated among idle clients. Dahlin et al. 10] describe an N Chance Forwarding algorithm for a cooperative cache in which the file caches of many client machines are coordinated to form a global cache. N Chance Forwarding tries to keep as many different blocks in global memory as ....

D. Comer and J. Griffioen. A New Design for Distributed Systems: The Remote Memory Model. In The Proceedings of the 1990 Summer USENIX Conference, pages 127--136. USENIX Association, June 1990.

No context found.

D. Comer and J. Gri#oen. A new design for distributed systems: the remote memory model. Proceedings of the USENIX 1991.

No context found.

D. Comer and J. Griffoen. A new design for distributed systems: the remote memory model. Proceedings of the USENIX 1991.

No context found.

D. Comer and J. Gri#oen. A new design for distributed systems: The remote memory model. In Proc. of the USENIX Summer Conference, pages 127--135, June 1990.

No context found.

Doug Comer and J. Griffioen. A New Design for Distributed Systems: The Remote Memory Model. In Proceedings of the 1990 Summer USENIX Conference (June 1990), 127135.

First 50 documents

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