Citations: Two-level replacement decisions in paging stores

16 citations found. Retrieving documents...

O. Babaoglu and D. Ferrari. Two-level replacement decisions in paging stores. IEEE Transactions on Computers, C-32(12):1151--1159, Dec. 1983.

Home/Search Document Not in Database Summary Related Articles Check

This paper is cited in the following contexts:

Flexible Reference Trace Reduction for VM Simulations - Scott Kaplan Amherst (2003) (2 citations) (Correct)

....values of k (10 to 100) are enough to allow OLR and SAD to achieve reduction factors of up to several orders of magnitude, while guaranteeing exact simulations. SAD and OLR are also useful when studying LRU approximations, such as CLOCK and SEGQ (segmented queue also known as hybrid FIFO LRU [BF83] or segmented FIFO [TL81] These replacement policies ignore the same high frequency referencing information that nearly any replacement policy will ignore, and that SAD and OLR discard from traces. This information is ignored not because these are LRU approximations, rather than pure LRU, but ....

....mechanism is designed so that CLOCK selects a page that has not been referenced recently. Any recently referenced page is likely to have at least one of its reference bits set. We also simulated the SEGQ replacement policy (segmented queue also known as two level replacement, hybrid FIFO LRU [BF83], or segmented FIFO [TL81] with the original and reduced traces. This replacement policy orders resident pages in two segments. The first segment is a FIFO queue that holds some fixed number of the most recently referenced pages. Pages evicted from the first level are inserted at the front of the ....

Ozalp Babaoglu and Domenico Ferrari. Two-level replacement decisions in paging stores. IEEE Trans- actions on Computers, C-32(12):1151-1159, December 1983.

Multicache-based Content Management for Web Caching - Cheng, Kambayashi (2000) (Correct)

....content management schemes for supporting implementation of high performance low overhead cache replacement policies, especially in the research of operating systems and database systems. For example, Segmented FIFO in [17] FBR in [14] Segmented LRU in [8] 2Q in [7] Ozalp Babaoglu et al. in [3] analyzed a family of two level content management schemes. Performance models are developed to measure these them. In Web caching, most caching schemes have a content management scheme to support efficient implementation. Least Relative Value (LRV) proposed by Luigi et al. [13] uses 10 FIFO ....

O. Babaoglu and D. Ferrari. Two-Level Replacement Decision in Paging Stores. IEEE Transactions on Computers, c-32(12):1151--1159, December 1983.

Optimal Trace Reduction for LRU-based Simulations - Yannis Smaragdakis University (Correct)

....blocks) our algorithm achieves reduction factors of up to several orders of magnitude. Our algorithm is also useful when studying approximations of LRU. The most prominent approximations, used in most actual operating systems, are clock and segq (segmented queue also known as hybrid FIFOLRU [BaFe83] or segmented FIFO [TuLe81] In [KSW98] we showed that the error introduced by our algorithm for approximations of LRU is small under 2 in number of faults in most cases. This error is smaller than that of stack deletion [Smit77] which is a common technique for removing high frequency ....

O. Babaoglu and D. Ferrari, "Two-Level Replacement Decisions in Paging Stores", IEEE Transactions on Computers, 32(12) (1983).

Buffer Replacement Using Online Optimization by Mining - Tung Kum Hoe (Correct)

....algorithm is First In First Out (FIFO) As the name implies, FIFO throw out the first page that goes into the buffer. Due to its simplicity, FIFO is usually the worst among all replacement algorithms. Among all the others existing buffer replacement algorithms, the Least Recently Used (LRU) [4, 9, 26] is the most popular among commercial systems. LRU replaces the buffer page that has not been accessed for the longest time. Besides LRU, other improved replacement algorithms are as follow : i) The CLOCK and GCLOCK Algorithms In the CLOCK algorithm, an use bit is added to every buffer page, ....

O. Babaoglu and D. Ferrai. Two-level replacement decisions in paging stores. IEEE Transactions on Computers, 32(12):1151-1159, Dec. 1993.

Compressed Caching and Modern Virtual Memory Simulation - Kaplan (1999) (5 citations) (Correct)

....[SKW99] Although these policies differ from LRU, their general behavior is similar to it, as the majority of references are to pages that all of these policies keep in an LRU queue. Policies like clock [Car84] and segmented queue (segq, also known as two level replacement or hybrid FIFO LRU ) BF83] represent the category LRU approximations. These two policies are the most used replacement policies in real systems. They make their eviction selections based on recency, although their heuristics to record recency are not as exact as an LRU queue. Because these policies are so similar to ....

....LRU, in contrast, can perform worse than EELRU by a factor proportional to the size of memory in the worst case a factor is usually in the thousands. This guaranteed property of EELRU is interesting both because of the ubiquity of LRU (and its approximations, e.g. segmented queue [TL81, BF83] and because of the commonality of the LRU worst case pattern (a simple, large loop) in practice. EELRU is designed to be timescale relative. Additionally, EELRU is firmly based on a distinct principle of program locality studies, that of timescale relativity (see Section 2.1 and [WKM94] ....

[Article contains additional citation context not shown here]

Ozalp Babaoglu and Domenico Ferrari. Two-level replacement decisions in paging stores. IEEE Transactions on Computers, C-32(12):1151--1159, December 1983.

EELRU: Simple and Effective Adaptive Page Replacement - Smaragdakis, Kaplan, Wilson (1999) (13 citations) (Correct)

....performance, regardless of system configuration. Getting good performance for all different applications and configurations is hard. Currently most operating systems use replacement algorithms that are approximations of Least Recently Used (LRU) replacement (e.g. segmented FIFO implementations [TuLe81, BaFe83]) Although very good in most cases, the performance of LRU based algorithms suffers for regular access patterns larger than the size of main memory. Such patterns are quite common in programs. For instance, a linear loop touching more memory pages than the memory capacity incurs faults for ....

....region l (late eviction point) MRU page) Figure 3: EELRU with WFL fallback: LRU axis and correspondence to memory locations. speculate that it is quite feasible. In particular, EELRU can be approximated using techniques identical to standard in kernel LRU approximations (e.g. segmented FIFO [TuLe81, BaFe83]) References to the most recently used pages do not matter for EELRU statistics and incur no overhead. Compared to LRU, the only extra requirement of EELRU is maintaining recency information even for pages that have been evicted. Since this information only changes at page fault time, the cost of ....

O. Babaoglu and D. Ferrari, "Two-Level Replacement Decisions in Paging Stores", IEEE Transactions on Computers, 32(12) (1983).

EELRU: Simple and Effective Adaptive Page Replacement - Smaragdakis, Kaplan, Wilson (1999) (13 citations) (Correct)

....LRU, in contrast, can perform worse than EELRU by a factor proportional to the number of memory pages in the worst case. This factor is usually in the thousands. This guaranteed property of EELRU is interesting both because of the ubiquity of LRU (and its approximations, e.g. segmented FIFO [TuLe81, BaFe83]) and because of the commonality of the LRU worst case pattern (a simple, large loop) in practice. Additionally, EELRU is firmly based on a distinct principle of program locality studies, that of timescale relativity (see also [WKM94] Program behavior can be studied at many timescales (for ....

....EELRU scheme allows for quite efficient implementations. Even though we have not provided an in kernel version of EELRU, we speculate that it is quite feasible. In particular, EELRU can be approximated using techniques identical to standard in kernel LRU approximations (e.g. segmented FIFO [TuLe81, BaFe83]) References to the most recently used pages do not matter for EELRU statistics and incur no overhead. Compared to LRU, the only extra requirement of EELRU is maintaining recency information even for pages that have been evicted. Since this information only changes at page fault time, the cost of ....

O. Babaoglu and D. Ferrari, "Two-Level Replacement Decisions in Paging Stores", IEEE Transactions on Computers, 32(12) (1983).

Trace Reduction for Virtual Memory Simulations - Scott Kaplan (3 citations) (Correct)

....to allow OLR and SAD to achieve reduction factors of up to several orders of magnitude, while guaranteeing exact simulations. SAD and OLR are also useful when studying approximations of LRU. The most prominent approximations are clock and segq (segmented queue also known as hybrid FIFO LRU [BaFe83] or segmented FIFO [TuLe81] These replacement policies ignore the same high frequency referencing information that nearly any replacement policy will ignore, and that SAD and OLR discard from traces. This information is ignored not because these are LRU approximations, but because references to ....

....chosen for eviction. This mechanism is designed so that clock selects a page that has not been referenced recently. Any recently referenced page is likely to have at least one of its reference bits set. We also simulated the segq replacement policy (segmented queue also known as hybrid FIFO LRU [BaFe83] or segmented FIFO [TuLe81] with the original and reduced traces. This replacement policy orders resident pages in two segments. The first segment is a FIFO queue that holds some fixed number of the most recently referenced pages. Pages evicted from the first level are inserted at the front of ....

O. Babaoglu and D. Ferrari, "Two-Level Replacement Decisions in Paging Stores", IEEE Transactions on Computers, 32(12) (1983).

Buffer Management in Distributed Database Systems: A Data.. - Lu (1998) (Correct)

....buffer frames are in use, which page from this private buffer should be replaced Among the above three issues, buffer replacement problem has been extensively studied. An early survey can be found in [9] The most popular and simple buffer replacement strategy is LRU, the Least Recently Used [15, 3]: When a new buffer frame is needed, the page in the buffer that has not been accessed for the longest time is replaced. LFU (Least Frequently Used) policy is another simple but effective buffer replacement policy based on the frequency a page is referenced [16] It relates a reference count to ....

O. Babaoglu and D. Ferrari. Two-level replacement decisions in paging stores. IEEE Transactions on Computers, 32(12):1151--1159, December 1983.

Automatic Heap Sizing: Taking Real Memory Into Account - Ting Yang Matthew (2004) (Correct)