J.G. Thompson. Efficient Analysis of Caching Systems. PhD thesis, University of California, Berkeley, 1987. 13  Home/Search   Document Not in Database   Summary   Related Articles   Check

....this algorithm is considerably faster the the general algorithm. Our study of this algorithm concentrates then on determining the sensitivity of performance to increasing B. The MIMD algorithms studied are parallelizations of the efficient serial LRU simulation algorithm de1 scribed in [7] and [10]. This algorithm works in the following manner. For every reference x t , let n(t) be the smallest index u t such that x t = xu , and let p(t) be the largest index v t such that x t = x v . We call n(t) and p(t) the next and previous occurrences of symbol x t . The serial algorithm we use ....

J.G. Thompson. Efficient Analysis of Caching Systems. PhD thesis, University of California, Berkeley, 1987. 13

.... with whole file caching, but could easily use partial file caching, or no caching (although a performance penalty would be paid) We chose to use whole file caching as most files in the assumed environment are small, and 76 of the files accessed are completely read or written [1] According to [20], whole file caching is a good method for files up to 100K, after which the benefits of transferring the whole file are reduced, as large files are not usually completely read or written during a file session. Our protocol can easily handle larger files by switching to a partial file caching ....

J. Thompson, "Efficient Analysis Of Caching Systems," Technical Report No. UCB/CSD 87/374, Computer Science Division, Univ. of California, Berkeley, Calif., Oct. 1987.

....size) separately, assuming LRU replacement. We now permit capacity misses, but as before no redundant transfers occur. Efficient single pass simultaneous simulation algorithms for this computation have long been available for the special case where document sizes and miss penalties are uniform [11, 33, 36] and have recently been generalized to non uniform sizes and miss costs [23] Using the Reeves Kelly algorithm we first compute browser cache hit rates for each client at every cache size. We then aggregate the results into a single success function for the entire client population. 4 To avoid ....

J. G. Thompson. Efficient analysis of caching systems. Technical Report UCB/CSD 87/374, U.C. Berkeley EECS, Oct. 1987.

.... A (s) Total miss cost over entire reference sequence ( cost cache size memory cost miss cost total cost optimal sizes Fig. 2. Cache costs as monotonic step functions. transformations of reference streams; replacement policies with this property are sometimes known as stack policies [11,30,32,37]. LRU and the variant of Perfect LFU that caches a requested document only if it has sufficiently high priority ( optional placement Perfect LFU ) are stack policies; FIFO and mandatory placement LFUs are not [30] 3 The first requirement is necessary because aggregate miss cost is monotonic ....

....per reference for a total of O(MN) time to process the entire sequence of M requests. For reasonable removal policies, however, it is possible to perform this computation in O(M log N) time and O(N) memory using an algorithm reminiscent of those developed for efficient processor memory simulation [11,32,37]; we describe our priority depth algorithm in Section 3.1. Given a pair (# t (x t ) t ) for each of M requests, we can compute A (s) after sorting these pairs on # in O(M log M) time and O(M) memory. This post processing sorting step is therefore the computational bottleneck for any trace ....

[Article contains additional citation context not shown here]

J.G. Thompson, Efficient analysis of caching systems. Technical Report UCB/CSD 87/374, Computer Science Division (EECS), University of California at Berkeley, October 1987. This is the author's PhD dissertation.

....memory access time. Reducing the cache miss rate is one factor for improving memory access performance. Cache misses occur for a number of reasons: cold start, capacity, and collisions [12] A number of cache line replacement algorithms have been proposed to reduce the number of cache collisions [2, 13, 17]. Instead of concentrating on cache organization we concentrate on the layout of a program on the memory space. Bershad et.al. suggested remapping cache addresses dynamically to avoid conflict misses in large direct mapped caches [3] An alternative approach is to perform code repositioning at ....

J.G. Thompson. Efficient analysis of caching systems. Ph.D. Dissertation, University of California, Berkeley, 1987. 22

....cache dimension: ae = N Gamma (R Gamma D) N ; 3.2) where D the total number of dimensional conflicts. For the example, ae = 8 Gamma (4 Gamma 2) 8 = 0:75. This is a general model and can be extended to account for other effects, such as conflicts due to multiprocessor cache coherence [32] and context switching (see Section 3.3) 27 3.1.2 Reference streams and cache dimensions A formal abstraction of a benchmark s trace is termed a reference stream. This is a sequence of references to addresses, w(k) of length N (0 k N ) When required, the addresses are represented by ....

J. G. Thompson, Efficient analysis of caching systems, Ph.D. dissertation, Computer Science Division, University of California, Berkeley, CA, Oct. 1987. Report No. UCB/CSD 87/374.

....the single pass simulation difficult. In this paper we propose a limited lookahead scheme to do single pass optimal cache simulation without the reverse pass. The sequential list based search of the stack is a bottleneck in stack simulation. This problem has been addressed for LRU replacement [3, 16, 22, 11]. These algorithms make use of the fact that the stack update is simple in LRU; the referenced entry is just moved from its current position to the top of the stack. Under OPT replacement, however, the relative positions of many entries other than the referenced entry can potentially change in the ....

....The tree based algorithms use splay trees [17] which are self balancing and well suited for cache simulation. In the LRU tree algorithm, the time of previous reference of an address is obtained by a hash lookup. The hit depth is then determined by a tree lookup that uses this time as the key[16, 22]. In the OPT tree implementation, a list of groups is maintained numbered in order. A descriptor is associated with each group that contains the first and last addresses of the group, and the number of entries in the group. All the addresses in the stack are maintained in in order form in one ....

J. G. Thompson. Efficient analysis of Caching Systems. PhD thesis, University of California, Berkeley, 1987.

....of program behavior [16,26] that helped design several good page replacement algorithms and caching strategies. In the 1980s, with the increasing trend towards distributed computing, the caching of files (located remotely) and the study of file reference behavior became an interesting topic [5,6,13,14,17,21,23,28,29]. Recently, we discovered that the frames on computer networks also exhibit locality behavior [9] The understanding of this behavior will help us design the large networks of the 1990s in an efficient manner. The trend toward networks becoming larger and faster, and addresses also increasing in ....

J. G. Thompson, "Efficient Analysis of Caching Systems," University of California, Berkeley, Report No. UCB/CSD 87/374.

....writing policy. However, results from four studies of UNIX timesharing traces can be used to help predict the best writing policy for clients and servers. In addition to the two previously mentioned studies by Ousterhout and Kent there are also studies that were done by Floyd [Flo86] and Thompson [Tho87]. Floyd s 15 studies are nearly identical to Ousterhout s studies so I will not mention them further. Thompson s study was a follow on study to the study done by Ousterhout et al. Thompson s results are based on very detailed traces of UNIX timesharing systems. The simplest policy for managing ....

....the file. In this case a client must be able to detect its stale data whenever it attempts to read data from the file. The amount of file sharing that occurs has an impact on the importance of cache consistency. Jim Thompson analyzed the amount of file sharing that occurred in a UNIX environment [Tho87] and got several interesting results: 20 Time C1 has file open for writing writing open for C2 has file Time reading open for C1 has file C1 has file open for reading writing open for C2 has file reading open for C1 has file Concurrent Write Sharing Sequential Write Sharing Figure 2 2. ....

[Article contains additional citation context not shown here]

J. G. Thompson, Efficient Analysis of Caching Systems, Phd Thesis, University of California at Berkeley, 1987.

....memory access time. Reducing the cache miss rate is one factor for improving memory access performance. Cache misses occur for a number of reasons: cold start, capacity, and collisions [13] A number of cache line replacement algorithms have been proposed to reduce the number of cache collisions [2, 14, 19]. Instead of concentrating on cache organization we concentrate on the layout of a program on the memory space. Bershad et.al. suggested remapping cache addresses dynamically to avoid conflict misses in large direct mapped caches [3] An alternative approach is to perform code repositioning at ....

J.G. Thompson. Efficient analysis of caching systems. Ph.D. Dissertation, University of California, Berkeley, 1987.

....[Baker et al. 1992] Additionally, the client that is attempting to access the new data must endure the latency of a synchronous disk write before it can begin writing. For many workloads, little write sharing occurs, so this delay in revoking ownership will have little effect on performance [Thompson, 1987, Baker et al. 1991, Blaze, 1993] Further, by maintaining consistency on a per block rather than a per file basis, the consistency protocol reduces ownership loss writes caused by false sharing. On the other hand, ownership loss writes could delay parallel workloads that have fine grained write ....

Thompson, J. (1987). Efficient Analysis of Caching Systems. PhD thesis, University of California at Berkeley.

....for deriving cache miss ratio, ae, for a given trace and a given cache dimension, ae = N Gamma (R Gamma D) N ; 3) where D the total number of dimensional conflicts. This is a general model and can be extended account for other effects, such as conflicts due to multiprocessor cache coherence [17] and context switching [16] The formal abstraction of a benchmark s trace is termed a reference stream. This is a sequence of address references, w(k) of length N (0 k N ) The addresses are addresses in the lowest level of a cache hierarchy, which is assumed to be a linear space (e.g. the ....

J. G. Thompson, Efficient analysis of caching systems. PhD thesis, Computer Science Division, University of California, Berkeley, California, Oct. 1987. Report No. UCB/CSD 87/374.

No context found.

Thompson87. J. Thompson, "Efficient Analysis of Caching Systems", PhD Thesis, 1987. University of California, Berkeley.

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