Principles of Optimal Page Replacement

Aho, Alfred V.; Denning, Peter J.; Ullman, Jeffrey D.

doi:10.1145/321623.321632

Cited by 240 publications

(107 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Another early proposal is the local page replacement policy [2], which forces a process to swap out one of its own pages when it requires more memory. Several operating systems implement this idea [8] as a way to isolate tasks performance.…”

Section: Thrashing Mitigationmentioning

confidence: 99%

See 1 more Smart Citation

Swap Fairness for Thrashing Mitigation

Goichon

Salagnac

Frénot

2013

Software Architecture

View full text Add to dashboard Cite

Abstract:The swap mechanism allows operating systems to manage more memory than the available RAM space, by temporarily storing unused memory pages on disk. However, disk transfers are way slower than normal RAM operations, and under memory pressure, the system may spend more time in retrieving and storing swapped pages than performing actual computation: this state is called memory thrashing. To reduce thrashing, several ideas were brought up to optimize page replacement algorithms and system-wide load. For instance, Linux currently implements the swap-token, a mechanism designed to immunize the memory pages of the heaviest process against swapping. Such a mechanism eliminates early thrashing peaks and improves general system performance if the process is to finish quickly. The swap-token may however be counterproductive when it is tricked into advantaging malicious or long-standing processes. This is particularily true in the context of shared hosting or virtualization, where multiple users run uncoordinated and selfish workloads. In this paper, we present an accounting layer that forces swap fairness among processes competing for main memory. It ensures that a process cannot perform longer swap operations than others, and delays the swap operations of processes abusing the swapping mechanism. With such a layer, we are able to significantly reduce the dispersion of execution times under memory pressure, and generally improve the performance of legit, memory-heavy processes running concurrently with abusive ones. Pour atténuer ce problème, les approches classiques proposent des algorithmes optimisés de remplacement de page, ou des techniques de réduction de la charge globale du système. Par exemple, Linux emploie une technique appelée jeton de swap (swap-token), conçue pour protéger l'un des processus du système contre tout déchargement de page. Ce mécanisme viseà permettreà ce processus de terminer son exécution le plus rapidement possible, de façonà faire redescendre la pression sur le swap. Cependant, le swap-token se révèle inefficace, voire contreproductif lorsqu'il se trompe et qu'il favorise un processus malveillant, ou simplement durable. Ce problème est particulièrement sévère dans le contexte de l'hébergement mutualisé ou de la virtualisation, où plusieurs utilisateurs doivent se partager une machine et y exécuter plusieurs charges de travail simulatnées etégoïste.Dans ce rapport, nous présentons une couche de facturation qui force l'équité entre tous les processus en compétition pour la mémoire. De cette façon, un processus ne peut pas s'accaparer le mécanisme de swap, ni retarder déraisonnablement les requêtes de swap des autres processus. Cette couche permet de réduire significativement la dispersion des temps d'exécution de programmes en situation de contention mémoire, et de manière générale, d'améliorer la performance des application légitimes lorsqu'elles s'exécutent en concurrence avec des applications abusives.

show abstract

Section: Thrashing Mitigationmentioning

confidence: 99%

“…Past work, such as local page replacement [2], load control [5] or the working set model [6], either minimize the concurrency or memory usage on the system, or are very expensive to implement properly. Jiang et al [10] propose a so-called token-ordered LRU policy meant to avoid early thrashing stages.…”

Section: Introductionmentioning

confidence: 99%

Swap Fairness for Thrashing Mitigation

Goichon

Salagnac

Frénot

2013

Software Architecture

View full text Add to dashboard Cite

show abstract

“…The independent reference model (IRM) [25,26] profiles an execution to determine the frequency with which each page in the working set is accessed. A synthetic stream is then generated that contains the same frequency for each page.…”

Section: Related Workmentioning

confidence: 99%

PSnAP: Accurate Synthetic Address Streams through Memory Profiles

Olschanowsky

Tikir

Carrington

et al. 2010

Languages and Compilers for Parallel Computing

View full text Add to dashboard Cite

Abstract. Memory address traces are an important information source; they drive memory simulations for performance modeling, systems design and application tuning. For long running applications, the direct use of an address trace is complicated by its size. Previous attempts to reduce trace size incurred a substantial penalty with respect to trace accuracy. We propose a novel method of memory profiling that enables the generation of highly accurate synthetic traces with space requirements typically under 1% of the original traces. We demonstrate the synthetic trace accuracy in terms of cache hit rates, spatial-temporal locality scores and locality surfaces. Simulated cache hit rates from synthetic traces are within 3.5% of observed and on average are within 1.0% for L1 cache. Our profiles are on average 60 times smaller than compressed traces. The combination of small profile sizes and high similarity to original traces makes our technique uniquely applicable to performance modeling and trace driven simulation of large-scale parallel scientific applications.

show abstract

“…LRU locality. To adapt to changes in the content and interests we use a LRU replacement policy [2]. Shortcuts that have been used recently receive a higher rank.…”

Section: Content Provider and Recommender Indexmentioning

confidence: 99%

Searching Dynamic Communities with Personal Indexes

Löser

Tempich²,

Quilitz

et al. 2005

The Semantic Web – ISWC 2005

View full text Add to dashboard Cite

Abstract. Often the challenge of finding relevant information is reduced to find the 'right' people who will answer our question. In this paper we present innovative algorithms called INGA (Interest-based Node Grouping Algorithms) which integrate personal routing indices into semantic query processing to boost performance. Similar to social networks peers in INGA cooperate to efficiently route queries for documents along adaptive shortcut-based overlays using only local, but semantically well chosen information. We propose active and passive shortcut creation strategies for index building and a novel algorithm to select the most promising content providers depending on each peer index with respect to the individual query. We quantify the benefit of our indexing strategy by extensive performance experiments in the SWAP simulation infrastructure. While obtaining high recall values compared to other state-of-the-art algorithms, we show that INGA improves recall and reduces the number of messages significantly.

show abstract

Principles of Optimal Page Replacement

Cited by 240 publications

References 14 publications

Swap Fairness for Thrashing Mitigation

Swap Fairness for Thrashing Mitigation

PSnAP: Accurate Synthetic Address Streams through Memory Profiles

Searching Dynamic Communities with Personal Indexes

Contact Info

Product

Resources

About