M.R. Korupolu and M. Dahlin, "Coordinated Placement and Replacement for Large-Scale Distributed Cached," IEEE Trans. Knowledge and Data Eng., vol. 13, 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in discrete location theory for several decades. For a survey of the area, we refer the reader to the texts by Mirchandani and Francis [38] and Love et al. 33] Problems in this area have also received attention recently due to their applicability to the problem of placing resources on a network [25, 26, 27]. As mentioned in Chapter 1, we focus on the metric versions of these problems; for recent work and pointers to the literature on the general (non metric) facility 12 location and k median problem, see [47] More generally, the text by Hochbaum [18] surveys the techniques used to obtain ....

M. Korupolu and M. Dahlin. Coordinated placement and replacement for large-scale distributed caches. In Proceedings of the IEEE Workshop on Internet Applications, pages 62--71, July 1999.

....large objects is quite rare. Moreover, download of such a huge object through WWW is unrealistic. 4. 3 Replacement Algorithms Dozens of object replacement algorithms for a cache are proposed, and developing better algorithms is still one of hot topics in the field of WWW caching (for example, see [16, 17, 18]) However, a better cache replacement algorithm is not the main motivation of this study, we adopt a classical LRU algorithm in our simulation, which is proved to achieve a moderate hit rate compared with other algorithms. 4.4 Failure Rates System failure can be characterized by two factors, ....

Madhukar R. Korupolu and Michael Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Proceedings of the IEEE Workshop on Internet Applications, San Jose, CA , USA, July 1999. IEEE.

....objects is quite rare. Moreover, download of such a huge object through WWW is unrealistic. 4.4. 3 Replacement Algorithms Dozens of object replacement algorithms for a cache are proposed, and developing better algorithms is still one of hot topics in the eld of WWW caching (for example, see [24, 25, 26]) However, a better cache replacement algorithm is not the main motivation of this study, we adopt a classical LRU algorithm in our simulation, which is proved to achieve a moderate hit rate compared with other algorithms. 4.4.4 Failure Rates System failure can be characterized by two factors, ....

Madhukar R. Korupolu and Michael Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Proceedings of the IEEE Workshop on Internet Applications, San Jose, CA , USA, July 1999. IEEE.

....here happens within a LAN that connects the cluster nodes, which is different from cooperative caching in another popular context where proxy servers scattering in a wide area network work cooperatively over long distance connections. The latter requires protocols that are more sophisticated [23,24,25]. The nodes in our cluster based server system work cooperatively, to fully utilize all the deployable physical memory in each node to form a cache layer for storing the site s contents. As handling objects in memory is many times faster than if they are on disk, the cluster nodes can respond ....

Korupolu M, Dahlin M. Coordinated Placement and Replacement for Large-Scale Distributed Caches. Proceedings of the 1999 IEEE Workshop on Internet Applications, San Jose, CA, USA, July 1999.

....to Workload and Network Characteristics Kang Won Lee Khalil Amiri Sambit Sahu Chitra Venkatramani IBM Thomas J. Watson Research Center Email: kangwon, amirik, sambits, chitrav us.ibm.com 1. INTRODUCTION A rich body of literature exists on several aspects of cooperative caching [1, 2, 3, 4, 5], including object placement and replacement algorithms [1] mechanisms for reducing the overhead of cooperation [2, 3] and the performance impact of cooperation [3, 4, 5] However, while several studies have focused on quantifying the performance benefit of cooperative caching, their conclusions ....

....Research Center Email: kangwon, amirik, sambits, chitrav us.ibm.com 1. INTRODUCTION A rich body of literature exists on several aspects of cooperative caching [1, 2, 3, 4, 5] including object placement and replacement algorithms [1] mechanisms for reducing the overhead of cooperation [2, 3], and the performance impact of cooperation [3, 4, 5] However, while several studies have focused on quantifying the performance benefit of cooperative caching, their conclusions on the effectiveness of such cooperation vary significantly. The source of this apparent disagreement lies mainly in ....

[Article contains additional citation context not shown here]

M. R. Korupolu and M. Dahlin, "Coordinated Placement and Replacement for Large-Scale Distributed Caches," in Proc. of the IEEE Workshop on Internet Applications, July 1999.

....most frequently accessed documents thereby avoiding wide area network transfers and reducing client download latencies. To further improve document retrieval performance for web clients, several researchers have investigated the potential for cooperation among closely located web proxy caches [1, 2, 3, 4, 5, 6, 7, 8, 9]. However, a closer examination of these studies reveals that there is no clear consensus as to whether cooperation among proxy caches results in any additional bene t. Our goal in this work is to provide a closer examination at various forms of cooperation that have been studied in the literature ....

....Access to such large rich media objects is expected to increase even further as broadband networks, currently reaching only 10 of the homes, extend their reach to the majority of residential areas. A large body of literature exists on the various aspects of cooperative caching in the Internet [1, 2, 3, 4, 5, 6, 7, 8, 9]. Rabinovich et al. 1] have focussed on cost aware algorithms for cooperative proxy caching, 1 while others [2, 3] have looked at the optimal object placement algorithm in a cooperative cluster and a content distribution network. Other researchers have focussed on mechanisms to reduce the ....

[Article contains additional citation context not shown here]

Madhukar R. Korupolu and Michael Dahlin, \Coordinated Placement and Replacement for Large-Scale Distributed Caches," in Proceedings of the IEEE Workshop on Internet Applications, pp. 62-71, July 1999.

No context found.

M. R. Korupolu and M. Dahlin. Coordinated placement and replacement for largescale distributed caches. In Proceedings of the IEEE Workshop on Internet Applications, pages 62--71, July 1999.

No context found.

M. Korupolu and M. Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Proc. of the 1999.

.... 19] there is reason to be concerned that caching alone will not signi cantly improve WAN service availability because much HTTP trac is uncachable [11, 44] This limitation motivates us to study the potential e ectiveness of other techniques such as hoarding [19] push based content distribution [13, 20], relaxed consistency, mobile extensions to ship service code to proxies or clients [2, 4, 9, 18, 37, 39, 45] anycast [3, 12, 45] and overlay routing [33] Although the performance bene ts of many of these techniques have been studied, their potential impact on end to end availability has not ....

....c adaptation to cope with disconnection by dynamically adjusting service semantics. In the context of web services, previous studies have examined the performance bene ts of caching [11, 36, 43] prefetching [10, 29, 22, 31] pushing updates [24, 36] push based content distribution [13, 20], server replication [26] mobile code [2, 4, 9, 18, 37, 39] and overlay routing [33] but the impact on end to end service availability of these techniques has not been systematically quanti ed. Systems implementing variations of some of these techniques have been built. The Netscape Navigator ....

[Article contains additional citation context not shown here]

M. Korupolu and M. Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Proc. of the 1999.

....little spare capacity, we observe that a threshold based prefetching scheme causes response times to increase by a factor of 2 due to interference, whereas prefetching using NPS decreases response times by 25 . 1 Introduction A number of studies have demonstrated the bene ts of web prefetching [12, 17, 24, 25, 32, 33, 42, 52]. And the attractiveness of prefetching appears likely to rise in the future as the falling prices of disk storage [14] and network bandwidth [41] make it increasingly attractive to trade increased consumption of these resources to improve response time and availability and thus reduce human wait ....

....interference. The damage is because TCP Nice is primarily designed for long ows. 8 Related Work Several studies have published promising results that suggest that prefetching (or pushing) content could signi cantly improve web cache hit rates by reducing compulsory and consistency misses [12, 17, 24, 25, 32, 33, 42, 52]. However, existing systems either su er from a lack of deployability or use threshold based magic numbers to address the problem of interference. Several existing commercial client side prefetching agents that require new code to be deployed to clients are available [39, 27, 53] At least one ....

M. Korupolu and M. Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Workshop On Internet Applications, June 1999.

....server trace, our system reduces the average demand request response times by 25 in comparison to a prefetching scheme that uses same prediction techniques and traditional techniques for limiting interference. 1 Introduction A number of studies have demonstrated the bene ts of Web prefetching [13, 18, 27, 28, 35, 36, 43, 53]. And the attractiveness of prefetching appears likely to rise in the future as the falling prices of disk storage [15] and network bandwidth [8, 42] make it increasingly attractive to trade increased disk storage and network bandwidth consumption for improved response time and availability. ....

....we demonstrated similar risks for poorly set thresholds with respect to the network. 8 Related Work Several studies have published promising results that suggest that prefetching (or pushing) content could signi cantly improve web cache hit rates by reducing compulsory and consistency misses [13, 18, 27, 28, 35, 36, 43, 53]. However, existing systems either su er from a lack of deployability or address the problem of interference using thresholdbased magic numbers. Several existing commercial client side prefetching agents that require new code to be deployed to clients are available [12, 30, 54] At least one ....

M. Korupolu and M. Dahlin. Coordinated placement and replacement for large-scale distributed caches. In Workshop On Internet Applications, June 1999.

....policies that take into account the di ering costs incurred in the retrieval of objects on cache misses. This has led to studies formulating generalizations of the traditional uniprocessor caching problems that account for the di ering costs [10, 19, 33] In a recent experimental study [21], Korupolu and Dahlin evaluate the practical performance of several placement and replacement algorithms for cooperative caching. Their simulation experiments demonstrate that, in practice, both our greedy placement algorithm as well as our amortizing placement algorithm are in fact very close to ....

M. Korupolu and M. Dahlin. Coordinated placement and replacement for large-scale distributed caches. In Proceedings of the IEEE Workshop on Internet Applications, pages 62-71, July 1999.

.... there is reason to be concerned that caching alone will not significantly improve WAN service availability because much HTTP trattlc is uncachable [11, 45] This limitation motivates us to study the potential effectiveness of other techniques such as hoarding [19] push based content distribution [13, 20], relaxed consistency, mobile extensions to ship service code to proxies or clients [2, 4, 9, 18, 37, 39, 46] anycast [3, 12, 46] and overlay routing [1, 33] Although the performance benefits of many of these techniques have been studied, their potential impact on end to end availability has ....

....adaptation to cope with disconnection by dynamically adjusting service semantics. In the context of web services, previous studies have examined the performance benefits of caching [11, 36, 44] prefetching [10, 29, 22, 31] pushing updates [24, 36] push based content distribution [13, 20], server replication [26] mobile code [2, 4, 9, 18, 37, 39] and overlay routing [33] but the impact on end to end service availability of these techniques has not been systematically quantified. Systems implementing variations of some of these techniques have been built. The Netscape Navigator ....

[Article contains additional citation context not shown here]

M. Korupolu and M. Dahlin. Coordinated Placement and Replacement for Large-Scale Distributed Caches. In Proc. of the 1999.

....replication systems also throws up questions of policy, namely object selection what to replicate and placement where to replicate. Our past research shows that by prefetching objects based on their popularity and lifetimes [20] significant improvements in hit rates may be obtained. In [13] and [18] we present algorithms to place objects in a distributed caching system so as to maximize hit rates in scenarios constrained by space and bandwidth respectively. In future, we intend to work on developing algorithms to place objects in more dynamic scenarios where both space and ....

Madhukar Korupolu and Mike Dahlin. Coordinated placement and replacement for large-scale distributed caches. In Workshop On Internet Applications, June 1999.

No context found.

M.R. Korupolu and M. Dahlin, "Coordinated Placement and Replacement for Large-Scale Distributed Cached," IEEE Trans. Knowledge and Data Eng., vol. 13, 2001.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated placement and replacement for large-scale distributed caches," IEEE Transactions on Knowledge and Data Engineering, vol. 14, no. 6, pp. 1317--1329, 2002.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated placement and replacement for large-scale distributed caches," IEEE Transactions on Knowledge and Data Engineering, vol. 14, no. 6, pp. 1317--1329, 2002.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated placement and replacement for large-scale distributed caches," IEEE Transactions on Knowledge and Data Engineering, vol. 14, no. 6, pp. 1317--1329, 2002.

No context found.

Madhukar R. Korupolu and Michael Dahlin, "Coordinated placement and replacement for large-scale distributed caches," in Proceedings of the IEEE Workshop on Internet Applications, June 1999, pp. 62--71.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated placement and replacement for large-scale distributed caches," in Proc. IEEE Workshop on Internet Applications, June 1999, pp. 62--71.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated placement and replacement for large-scale distributed caches," IEEE Transactions on Knowledge and Data Engineering, vol. 14, no. 6, pp. 1317--1329, November/December 2002.

No context found.

Madhukar R. Korupolu and Michael Dahlin, "Coordinated placement and replacement for large-scale distributed caches," in Proceedings of the IEEE Workshop on Internet Applications, June 1999, pp. 62--71.

No context found.

M. R. Korupolu and M. Dahlin, "Coordinated Placement and Replacement for Large-Scale Distributed Caches," in IEEE Workshop on Internet Applications, July 1999.

No context found.

M. Korupolu and M. Dahlin. Coordinated placement and replacement for large-scale distributed caches. In Proc. IEEE Workshop on Internet Applications, pages 62--71, July 1999.

No context found.

M.R. Korupolu, and M. Dahlin. \Coordinated Placement and Replacement for Large-Scale Distributed Caches," Proceedings of IEEE Workshop on Internet Applications, pages 62-71, July 1999.

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