K. Li and K. Petersen, Evaluation of memory system extensions, in Proc. 18-th International Symposium on Comp. Arch., 1991, pp. 84--93.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....scheduling can thus be viewed as an instance of gang scheduling of memory resources, in this case DRAM pages. Such explicit, cooperative management of shared resources has been shown to be an important factor in obtaining good performance on multiprocessor platforms. For instance, Li and Petersen [LiP91] show that for memory system extensions, direct management of remote memories performs better than using the extended memory modules as a transparent cache between main memory and disk. Leutenegger [Leu90] and Ousterhout et al. Ous80] argue for gang scheduling of CPU resources. Burger et al. ....

K. Li and K. Petersen, "Evaluation of Memory System Extensions", Proceedings of the 18th Annual International Symposium on Computer Architecture (ISCA), published as ACM SIGARCH Computer Architecture News, 19(3):84-93, May 1991.

....an essential part of justifying taking away from the operating system a resource that would otherwise be used to improve its performance. Hidden memory, with only some kind of client server interface for getting to it, is bound to be slower than regular memory. This was observed by Li and Petersen [12]. They found that segregating a large, slow portion of memory and using the small, fast memory as a cache gave poorer performance than using the slower memory directly. Since we are segregating memory of equal performance, we expect an outright performance loss on many applications. On the other ....

....by VM 370 since they were not available on the prototype hardware. This additional protection would eliminate the possibility of an operating system failure in boot or address translation service damaging durable (hidden) memory. This sharing of main memory was also examined by Li and Petersen [12]. They observed that using a slower, extended memory as a cache (as a stage of memory between fast main memory and disk) gave poorer overall system performance than accessing it directly (as the bulk of a much larger but slower main memory) for applications with large data structures. However, ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proceedings of the 18th Annual International Symposium on Computer Architecture, pages 84--93, May 1991.

....4 presents the implementation of the pager as a device driver. Section 5 presents our performance results which are very encouraging. Section 6 presents some aspects that we plan to explore as part of our future work. Finally, section 7 presents our conclusions. 2 Related Work Li and Petersen [10] have implemented a related system where they add main memory module on the I O bus (VME bus) of a computer system. This memory module can be used both as backing store, and as (slow) main memory accessed via simple load and store operations. Although this approach increases the amount of memory ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proc. 18-th International Symposium on Comp. Arch., pages 84--93, 1991.

....remote memory to commit transaction writes. These implementations generally target different environments (mobile computing, file systems, DBMSes) and do not attempt to take advantage of application specific information to reduce the cost of paging in parallel applications. Li and and Petersen [10] have implemented a system where they add memory modules on the I O bus (VME bus) of a computer system. This memory can be used both as backing store (another level in the memory hierarchy) and as slower main memory accessed via simple load and store operations. Although this approach allows the ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proceedings of the Eighteenth International Symposium on Computer Architecture, pages 84--93, Toronto, Canada, May 1991.

....has been previously explored in the literature. For example, several file systems [2, 6, 16, 22] use the collective main memory of several clients and servers as a large file system cache. Paging systems may also use remote main memory in a workstation cluster to improve application performance [12, 18, 21, 24]. Even Distributed Shared Memory systems can exploit the remote main memory in a NOW [11, 7] for increased performance and reliability. For example, Feeley et. al describe a log based coherent system that integrates coherency support with recoverability of persistent data [11] Their objective is ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proc. 18-th International Symposium on Comp. Arch., pages 84--93, 1991.

....is better than REMOTE MAIN with the exception of the TRANS application, where the system serviced an extraordinary amount of page faults. REMOTE MAIN is almost always the worst policy, because the remote memory access cost is large (100 to 1) over the local memory access cost. Li and Petersen [12] have done experiments on a system where the remote memory access cost was only twice the local memory access cost, and in their system, REMOTE MAIN was much more attractive than RAM DISK. Unfortunately, as architecture trends 2 The reader may notice that all policies have a somewhat high ....

....that uses remote memory both as main memory and as backing store, ii) the evaluation of memory management policies for real applications, and (iii) the systematic exploration of various parameters that influence the use of remote memory for storing an application s data. Li and Petersen [12] have implemented a related system where they add main memory on the I O bus (VME bus) of a computer system. This memory can be used both as backing store, but also as (slow) main memory accessed via regular load and store operations. Several research groups study the issues in using remote memory ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proc. 18-th International Symposium on Comp. Arch., pages 84--93, 1991.

....suggest that REX results in noticeable performance improvement over the unmodified EXODUS storage manager. 4 Related Work Using Remote Main Memory to improve the performance and reliability of I O in a NOW has been previously explored in the literature in file systems [3, 4, 5] pagers [13, 14, 6], even Distributed Shared Memory systems [15] The closest to our research is the Harp file system [5] Harp uses replicated file servers to tolerate single server failure. Each file server is equipped with a UPS to tolerate power failures, and speedup synchronous write operations. Although we use ....

K. Li and K. Petersen. Evaluation of memory system extensions. In ISCA, 1991.

....has been previously explored in the literature. For example, several file systems [1, 8, 18, 21] use the collective main memory of several clients and servers as a large file system cache. Paging systems may also use remote main memory in a workstation cluster to improve application performance [14, 19, 20, 22]. Even Distributed Shared Memory systems, can exploit the remote main memory in a NOW [13, 9] for increased performance and reliability. The closest of these systems to our research is the Harp file system [21] Harp uses replicated file servers to tolerate single server failure. Each file server ....

K. Li and K. Petersen. Evaluation of Memory System Extensions. In Proc. 18-th International Symposium on Comp. Arch., pages 84--93, 1991.

....insignificant, showing that lightly loaded computation nodes can also be memory servers efficiently. 5 Related Work Memory hierarchies and virtual memory systems have been studied in great detail. Earlier work focused on uniprocessor architectures, with emphasis on multilevel memory hierarchies [2, 25, 32, 23], memory access patterns [1] virtual memory systems [10, 11] memory and disk caches [33, 34] Most related work on memory management for parallel architectures includes page placement strategies for distributed shared memory architectures, and shared virtual memory systems. Non uniform memory ....

Kai Li and Karin Petersen. Evaluation of Memory System Extensions. In Proceedings of the 18th International Symposium on Computer Architecture, pages 84--93, May 1991.

No context found.

K. Li and K. Petersen, Evaluation of memory system extensions, in Proc. 18-th International Symposium on Comp. Arch., 1991, pp. 84--93.

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