N. Young. On-line file caching. In Proc. of the 9th Annual ACM-SIAM Symp. on Discrete Algorithms, pages 82--86, 1998. 266  Home/Search   Document Details and Download   Summary   Related Articles   Check

....that varies with different network RTTs. Under varying establishment costs, the objective is to optimize trade offs of total open time and combined connection establishment costs. An analogous extension for document caching is when documents have varying fetching costs. In this context, Young [37, 38] proposed a generalization of lru termed GreedyDual. An experimental study on Web proxy traces was performed in [9] GreedyDual generalizes lru and can be adopted for connection management. The mpg policy extends to handle varying establishment costs [10] Thus, our attribute based policies can be ....

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998.

....problems The plain LRU, LFU, and LRFU algorithms are designed for pages with equal sizes and miss costs. Web objects, however, vary significantly in both size and cost of a miss. This motivated the development of cache replacement algorithms that account for varying page sizes and fetching costs [5, 8, 3, 4]. Experiments showed that the optimally competitive Landlord algorithm performs well on Web caching sequences [8, 3] Other experiments, however, show that it can be outperformed by perfect LFU [2] A natural open problem is thus to extend our results to this more general model. That is, develop ....

....vary significantly in both size and cost of a miss. This motivated the development of cache replacement algorithms that account for varying page sizes and fetching costs [5, 8, 3, 4] Experiments showed that the optimally competitive Landlord algorithm performs well on Web caching sequences [8, 3]. Other experiments, however, show that it can be outperformed by perfect LFU [2] A natural open problem is thus to extend our results to this more general model. That is, develop natural hybrid policies of Landlord and weighted LFU. ....

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998. 8

....haimk math.tau.ac.il September 14, 2001 Abstract We give an integer programming formulation of the paging problem with varying sizes and fetching costs. We use this formulation to provide an alternative proof that a variant of the algorithm greedy dual size previously considered in [4, 15] is k Gammah 1 competitive against the optimal strategy with cache size h k. Our proof provides further insights to greedy dual size. We also indicate how the same integer programming formulation has been recently used [11, 1] to obtain approximation algorithms to the NP complete offline ....

....offline strategy that uses a cache of size h k. Cao and Irani proved that greedy dual size is k competitive against an optimal strategy that also uses a cache of size k. Independently, Young An abstract of this paper appeared in Proc. 10th ACM SIAM Symposium on Discrete Algorithms, 1999 [5] [15] studied a slightly different variant of greedy dual size that he called the landlord algorithm. Young proved [15] that landlord is k= k Gamma h 1) competitive against an optimal offline strategy that uses a cache of size h for every h k. Young s proof that greedy dual is k= k Gamma h 1) ....

[Article contains additional citation context not shown here]

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998. 9

....upon partitioning the cache according to replacement cost. 2.1 Existing Cost Aware Algorithms The theoretical community has studied cost aware algorithms as k server problems [12, 20] A restricted class of k server problems, weighted caching, is closely related to cost aware caching. LANDLORD [40] is a significant algorithm in the literature, which we use for comparison. LANDLORD is closely related to the leading web caching algorithm [10] LANDLORD combines replacement cost, cache object size, and locality by extending both LRU and FIFO to include cost and variable cache object sizes ....

N. E. Young. On-line File Caching. In Proceedings of the Ninth Annual ACM-SIAM Symposium on Discrete Algorithms, Balitmore, MD, January 17-19, 1999.

....multiple times by the same user over a very short period. # Shared Access to Files: The same file after it is read into cache, is also referenced by different users. Some earlier works on file caching in distributed systems and the staging of files from tertiary storage have been presented in [8, 10, 14, 15, 16, 18]. Recent studies on caching have focused more on webcaching [1, 3, 5, 12, 17] Cao and Irani [3] present a relative comparison of various cache replacement policies that have been proposed for web caching. Their work discusses some of the merits and concerns 3 Network Site D HRM HRM ....

....Use of a Storage Resource Manager in a Data Grid of replacements policies such as Least Frequently Used (LFU) Least Recently Used (LRU) etc. They propose a replacement policy for web caching called the Greedy Dual Size (GDS) 3] It is a variant of the replacement policy termed Greedy Dual (GD) [18], that was originally proposed for main memory caching of fixed size pages. It is also perceived as a generalization of the LRU when there is some variability in the cost of reading pages from secondary storage. We have included an implementation of the GDS policy with file transfer and processing ....

N. Young. On-line file caching. In SODA: ACM-SIAM Symposium on Discrete Algorithms (A Conference on Theoretical and Experimental Analysis of Discrete Algorithms), 1998. 15

....achieves the optimal max max ratio, but is not competitive at all. Moreover, strongly competitive randomized algorithms (PARTITION [21] and EQUITABLE [1] appear to be too complex to be implemented in hardware. On the other hand, Young [26] proposed loose competitiveness. In particular, Young [27] showed that if an algorithm A k (k h 1) competes with OPTn, then for most choices of k, the total cost of A on sequence R is insignificant, or A c competes with OPT on sequence R for a constant c. We note that ratio c depends on what retrieval cost is considered insignificant and the ....

Neal E. Young. On-line file caching. In Proceedings of the 9th Annual ACMSIAM Symposium on Discrete Algorithms (SODA-98), pages 82-86, New York, January 25-27 1998. ACM Press.

....upon partitioning the cache according to replacement cost. 2.1 Existing Cost Aware Algorithms The theoretical community has studied cost aware algorithms as k server problems [12, 20] A restricted class of k server problems, weighted caching, is closely related to cost aware caching. LANDLORD [40] is a significant algorithm in the literature, which we use for comparison. LANDLORD is closely related to the leading web caching algorithm [10] LANDLORD combines replacement cost, cache object size, and locality by extending both LRU and FIFO to include cost and variable cache object sizes ....

N. E. Young. On-line File Caching. In Proceedings of the Ninth Annual ACM-SIAM Symposium on Discrete Algorithms, Balitmore, MD, January 17-19,

....The content hit rate is measured per object or per byte and sometimes weighted by estimated object fetching cost. It is dictated by the available cache storage and the replacement policy used. Replacement policies for Web caches were extensively studied (e.g. 2] 3] 4] 5] 6] 7] 8] [9]) Policies that seem to perform well are Least Recently Used (LRU, which evicts the least recently requested object when the cache is full) Least Frequently Used (LFU, which evicts the least frequently requested object) and GreedyDual Size (which accounts for varying object sizes and fetching ....

N. Young, "On line file caching," in Proc. 9th ACM-SIAM Symposium on Discrete Algorithms, ACM-SIAM, 1998.

....not get such notifications. Our goal is to minimize the total activation cost incurred while servicing the sequence oe. We consider both on line and off line scenarios as well as distributed and centralized ones. 2. 1 Varying establishment costs and buffer sizes As in the case of a single cache [10, 21, 22], the connection caching model can be generalized by allowing varying sizes and fetching costs. In the TCP context, sizes may correspond to socket buffers allocations, but use of uniform sizes seems most reasonable. Establishment costs may capture the user perceived latency caused by ....

N. Young. On line file caching. In Proc. 9th ACMSIAM Symposium on Discrete Algorithms. ACMSIAM, 1998.

....environment exhibit different characteristics than in traditional operating systems paging. One wellrecognized difference is the variability in sizes and fetching costs of different resources (pages) and indeed, replacement policies that incorporate these parameters were suggested and evaluated [7, 13, 21, 20]. As is the case in traditional paging, individual requests remain hard to predict. We focus on two separate predictable aspects of the traffic in the web environment. The first is the usage levels experienced by large proxy and (typically) web servers. The second is distinct per page ....

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998.

....that may occur when requests and their responses are routed within the network. We also remark that cost models have been adopted in uniprocessor caching systems to model scenarios in which the costs incurred in the retrieval of objects on cache misses may vary from one object to another [10, 19, 33]. With regard to uniprocessor caching schemes, recent research has addressed the challenge of designing cache replacement policies that take into account the differing costs incurred in the retrieval of objects on cache misses. This has led to studies formulating generalizations of the ....

....addressed the challenge of designing cache replacement policies that take into account the differing 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 differing 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 ....

N. E. Young. On-line file caching. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms, pages 82--86, January 1998. 29

....USA fedith,hklg research.att.com October 27, 1998 Abstract We give an integer programming formulation of the paging problem with varying sizes and fetching costs. We use this formulation to provide an alternative proof that a variant of the algorithm greedy dual size previously considered in [2, 7] is k 1 k Gammah 1 competitive against the optimal strategy with cache size h k. Our proof provides further insights to greedy dual size. We further exploit the linear programming relaxation to obtain an approximate solution to the NP complete offline problem. 1 Introduction We consider ....

....fetching costs are uniform. Young proved that greedy dual is k= k Gamma h 1) competitive against an optimal offline strategy that uses cache of size h k. Cao and Irani proved that greedy dual size is k competitive against an optimal strategy that also uses cache of size k. More recently, Young [7] extended the result of Cao and Irani and proved that greedy dual size is k= k Gamma h 1) competitive against an optimal offline strategy that uses cache of size h for every h k. An abstract of this paper will appear in Proc. 10th ACM SIAM Symposium on Discrete Algorithms, 1999. 1 Young s ....

[Article contains additional citation context not shown here]

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998. 7

....objects against the improved hit rate achieved by reducing replication and storing more unique objects. In this work, we examine an optimal placement algorithm and three practical placement algorithms and compare them to several uncoordinated replacement algorithms (such as LFU, LRU, greedy dual [2, 23]) and a novel coordinated replacement algorithm. We drive this comparison with simulation based on both synthetic and trace workloads. The synthetic workloads allow us to examine system behavior in a wide range of situations, and the trace allows us to examine performance under a workload of ....

....algorithm evicts the least recently used object. LFU replacement. When a cache miss occurs, this algorithm evicts the object with the least (local) frequency of access. Greedy dual replacement. This is a generalization of the LRU algorithm to the case where each object has a different fetch cost [2, 23]. The motivation behind the greedy dual algorithm is that the objects with larger fetch costs should stay in the cache for a longer time. The algorithm maintains a value for each object that is currently in the cache. When an object is fetched into the cache, its value is set to its fetch cost. ....

[Article contains additional citation context not shown here]

N. E. Young. On-line file caching. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms, pages 82--86, January 1998.

....Web caching is the temporary local storage of WWW pages by a browser for later retrieval. From the user s point of view, the primary benefit of caching is reduced latency, as the time to access locally stored objects is minimal. We adopt the following standard general model of web caching [1, 5, 8, 14]: Web Caching Problem Statement: The browser is given an online sequence S of page requests, where each page p i 2 S has a size s(i) say, in bytes) and an access time t(i) that is required if p i is not cached. If the requested page p i is not in the cache (this is called a cache miss) then the ....

....results show that for both Web Caching and Site Search, an online algorithm needs a logarithmically sized cache to be constant competitive. In section 5, we analyze the online algorithm Landlord (this algorithm is a generalization of LRU and is also called Greedy Dual Size in the literature) [3, 5, 14]. Although we will state all results in the cost model, the results hold for the general model if k is replaced by k L . We show that Landlord is O i min i k; log n k j n 1= k 1) j competitive for Web Caching on depth first tree traversal sequences. We also show that the online ....

[Article contains additional citation context not shown here]

N. Young, "On-line file caching", ACM/SIAM Symposium on Discrete Algorithms, 82--86, 1998. 16

....of popular objects against the improved hit rate by reducing replication and storing more unique objects. In this work, we examine an optimal placement algorithm and three practical placement algorithms and compare them to several uncoordinated replacement algorithms (such as LFU, LRU, GreedyDual [2, 23]) and a novel coordinated replacement algorithm. We drive this comparison with simulation based on both synthetic and trace workloads. The synthetic workloads allow us to examine system behavior in a wide range of situations, and the trace allows us to examine performance under a workload of ....

....evicts the least recently used object. LFU Replacement. When a cache miss occurs, this algorithm evicts the object with the least (local) frequency of access. GreedyDual Replacement. This is a generalization of the LRU algorithm to the case where each object has a different but fixed miss cost [2, 23]. The motivation behind the GreedyDual algorithm is that the objects with larger cost should stay in the cache for a longer time. The algorithm maintains a value for each object that is currently in the cache. When an object is fetched into the cache, its value is set to its fetch cost. When a ....

[Article contains additional citation context not shown here]

N. E. Young. On-line file caching. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms, pages 82--86, January 1998.

....extensively in the context of processor architecture and file systems, the Web introduces more complicated challenges, since resources vary substantially in their sizes, fetching costs, and access patterns. Indeed, new replacement policies have been developed to incorporate these parameters [1, 2, 3, 4]. Although the Web exhibits high variability between the access patterns of different resources, individual resources have more regular acccess patterns. The most important factor for a good replacement policy remains predicting the next request time of a resource which depends on the ability to ....

N. Young, "On line file caching," in Proc. 9th ACM-SIAM Symposium on Discrete Algorithms, ACM-SIAM, 1998.

....extensively in the context of processor architecture and file systems, the Web introduces more complicated challenges, since resources vary substantially in their sizes, fetching costs, and access patterns. Indeed, new replacement policies have been developed to incorporate these parameters [1, 2, 3, 4]. Although the Web exhibits high variability between the 1 Revision of an ESA 98 paper access patterns of different resources, individual resources have more regular acccess patterns. The most important factor for a good replacement policy remains predicting the next request time of a ....

N. Young, "On line file caching," in Proc. 9th ACM-SIAM Symposium on Discrete Algorithms, ACM-SIAM, 1998.

....algorithms determine which objects are to be evicted when a cache miss occurs. We examine an optimal placement algorithm and three practical placement algorithms and compare them to several local, uncoordinated replacement algorithms and a new hierarchical extension to the GreedyDual algorithm [3, 25]. We drive this comparison with simulation studies based on both synthetic and trace workloads. The synthetic workloads allow us to examine system behavior in a wide range of situations, and the trace allows us to examine performance under a workload of widespread interest: web browsing. Based on ....

....displacing a more frequently used object. This strategy too works best when the accesses are drawn from a fixed probability distribution and are uncorrelated. GreedyDual Replacement. This is a generalization of the LRU algorithm to the case where each object has a different but fixed miss cost [3, 25]. The motivation behind the GreedyDual algorithm is that the objects with larger cost should stay in the cache for a longer time. The algorithm maintains a value for each object that is currently in the cache. When an object is fetched into the cache, its value is set to its fetch cost. When a ....

[Article contains additional citation context not shown here]

N. E. Young. On-line file caching. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms, pages 82--86, January 1998.

No context found.

N. Young. On-line file caching. In Proc. of the 9th Annual ACM-SIAM Symp. on Discrete Algorithms, pages 82--86, 1998. 266

No context found.

N. Young. On-line file caching. In Proc. of the 9th Annual ACM-SIAM Symp. on Discrete Algorithms, pages 82--86, 1998. 266

No context found.

N. Young. On line file caching. In Proc. 9th ACM-SIAM Symposium on Discrete Algorithms. ACM-SIAM, 1998.

No context found.

N. E. Young. On-line file caching. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms, pages 82--86, January 1998.

No context found.

N. E. Young. On-line file caching. Algorithmica, 33:371--383, 2002.

No context found.

N. Young. On-line file caching. Algorithmica, 33(1):371--383, 2002. 15

No context found.

Neal E. Young. On-line file caching. Algorithmica, 33(3):371--383, 2002. 20

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