Assaf Schuster | Technion Israel Institute of Technology (original) (raw)
Papers by Assaf Schuster
Communications of The ACM, Jul 1, 2014
Proceedings of the 13th ACM International Systems and Storage Conference, 2020
ACM SIGPLAN Notices, 2014
The number of guest virtual machines that can be consolidated on one physical host is typically l... more The number of guest virtual machines that can be consolidated on one physical host is typically limited by the memory size, motivating memory overcommitment. Guests are given a choice to either install a "balloon" driver to coordinate the overcommitment activity, or to experience degraded performance due to uncooperative swapping. Ballooning, however, is not a complete solution, as hosts must still fall back on uncooperative swapping in various circumstances. Additionally, ballooning takes time to accommodate change, and so guests might experience degraded performance under changing conditions. Our goal is to improve the performance of hosts when they fall back on uncooperative swapping and/or operate under changing load conditions. We carefully isolate and characterize the causes for the associated poor performance, which include various types of superfluous swap operations, decayed swap file sequentiality, and ineffective prefetch decisions upon page faults. We address t...
Proceedings of the 25th Symposium on Operating Systems Principles, 2015
There are several reconfiguring models of parallel computation that are considered in the publish... more There are several reconfiguring models of parallel computation that are considered in the published literature, depending on their switching capabilities. Three of the more popular ones are the Horizontal-Vertical Reconfigurable Mesh (HV-RN model) [LM89a, LM89b, ML89], the Linear Reconfigurable Mesh (LRN model) [MPRS87, BPRS91], and the General Reconfigurable Mesh (RN model) [Wan91, Sch91]. Can these reconfigurable models (in particular two dimensional arrays) be the basis for the design of massively parallel computers? Perhaps the most basic related question is: given an algorithm which is designed for a large reconfigurable mesh, can it be executed efficiently on a smaller reconfigurable mesh? We call it the simulation problem. Despite the fact that machines already exist and a large number of efficient algorithms are known for reconfigurable arrays, this question was left open for all reconfiguring models of computation. In this work we give several positive answers to the simula...
Jpdc, 1998
This paper concerns the computational aspects of the reconfigurable network model. The computatio... more This paper concerns the computational aspects of the reconfigurable network model. The computational power of the model is investigated under several network topologies and assuming several variants of the model. In particular, it is shown that there are reconfigurable machines based on simple network topologies, that are capable of solving large classes of problems in constant time. These classes depend on the kinds of switches assumed for the network nodes. Reconfigurable networks are also compared with various other models of parallel computation, like PRAM's and Branching Programs. Part of this work is to be presented at the 18th International Colloquium on Automata, Languages, and Programming (ICALP), July 1991, Madrid. y Department of Computer Science, The Hebrew University, Jerusalem 91904, Israel. E-mail: yosi@humus.huji.ac.il, Supported by Eshcol Fellowship. z Department of Applied Mathematics and Computer Science, The Weizmann Institute, Rehovot 76100, Israel. E-mail: p...
Lecture Notes in Computer Science, 2007
This paper focuses on local computations of distributed aggregation problems on fixed graphs. We ... more This paper focuses on local computations of distributed aggregation problems on fixed graphs. We define a new metric on problem instances, Veracity Radius (VR), which captures the inherent possibility to compute them locally. We prove that VR yields a tight lower bound on output-stabilization time, i.e., the time until all nodes fix their outputs, as well as a lower bound on quiescence time. We present an efficient aggregation algorithm, I-LEAG, which reaches both output stabilization and quiescence within a time that is proportional to the VR of the problem instance, and is also efficient in terms of per-node communication and memory. We empirically show that the VR metric also effectively captures the performance of previously suggested efficient aggregation protocols, and that I-LEAG significantly outperforms these protocols in several respects. Categories and Subject Descriptors C.4 [Performance of Systems]: [Performance attributes];
Proceedings of the 5th Jerusalem Conference on Information Technology, 1990. 'Next Decade in Information Technology', 1990
ABSTRACT
International Journal of Parallel Programming, 2006
Case-based reasoning (CBR) systems rely on the conceptual ordering of entities called cases. If a... more Case-based reasoning (CBR) systems rely on the conceptual ordering of entities called cases. If atomic case features are allowed to assume numeric as well as symbolic values, then a systematic comparison regime is needed to aggregate similarity scores. A common approach to deal with real-numbered features is normalisation. However, there are two conspicuous problems with this procedure: the similarity between two features is dependent on the corresponding values of all other cases to be ranked; and real-numbered features are often interpreted by human experts according to conceptual constraints associated with features. In such situations, a conceptual distance between two features should be determined rather than the length of a `gap' on a linear scale. Within the framework of a comprehensive case-knowledge architecture, the notion of a concept frame that can be associated with a case feature is proposed. Through this component it is possible to represent polymorphic atomic cas...
Proceedings Second International Workshop on High-Level Parallel Programming Models and Supportive Environments, 1997
Many attempts are made to integrate the resources and services of distributed computational envir... more Many attempts are made to integrate the resources and services of distributed computational environments into vir-tual parallel machines (VPMs). While utilizing idle CPU cycles and thus being very cheap and available to everyone, such metacomputing environments will ...
Lecture Notes in Computer Science, 2007
Communications of The ACM, Jul 1, 2014
Proceedings of the 13th ACM International Systems and Storage Conference, 2020
ACM SIGPLAN Notices, 2014
The number of guest virtual machines that can be consolidated on one physical host is typically l... more The number of guest virtual machines that can be consolidated on one physical host is typically limited by the memory size, motivating memory overcommitment. Guests are given a choice to either install a "balloon" driver to coordinate the overcommitment activity, or to experience degraded performance due to uncooperative swapping. Ballooning, however, is not a complete solution, as hosts must still fall back on uncooperative swapping in various circumstances. Additionally, ballooning takes time to accommodate change, and so guests might experience degraded performance under changing conditions. Our goal is to improve the performance of hosts when they fall back on uncooperative swapping and/or operate under changing load conditions. We carefully isolate and characterize the causes for the associated poor performance, which include various types of superfluous swap operations, decayed swap file sequentiality, and ineffective prefetch decisions upon page faults. We address t...
Proceedings of the 25th Symposium on Operating Systems Principles, 2015
There are several reconfiguring models of parallel computation that are considered in the publish... more There are several reconfiguring models of parallel computation that are considered in the published literature, depending on their switching capabilities. Three of the more popular ones are the Horizontal-Vertical Reconfigurable Mesh (HV-RN model) [LM89a, LM89b, ML89], the Linear Reconfigurable Mesh (LRN model) [MPRS87, BPRS91], and the General Reconfigurable Mesh (RN model) [Wan91, Sch91]. Can these reconfigurable models (in particular two dimensional arrays) be the basis for the design of massively parallel computers? Perhaps the most basic related question is: given an algorithm which is designed for a large reconfigurable mesh, can it be executed efficiently on a smaller reconfigurable mesh? We call it the simulation problem. Despite the fact that machines already exist and a large number of efficient algorithms are known for reconfigurable arrays, this question was left open for all reconfiguring models of computation. In this work we give several positive answers to the simula...
Jpdc, 1998
This paper concerns the computational aspects of the reconfigurable network model. The computatio... more This paper concerns the computational aspects of the reconfigurable network model. The computational power of the model is investigated under several network topologies and assuming several variants of the model. In particular, it is shown that there are reconfigurable machines based on simple network topologies, that are capable of solving large classes of problems in constant time. These classes depend on the kinds of switches assumed for the network nodes. Reconfigurable networks are also compared with various other models of parallel computation, like PRAM's and Branching Programs. Part of this work is to be presented at the 18th International Colloquium on Automata, Languages, and Programming (ICALP), July 1991, Madrid. y Department of Computer Science, The Hebrew University, Jerusalem 91904, Israel. E-mail: yosi@humus.huji.ac.il, Supported by Eshcol Fellowship. z Department of Applied Mathematics and Computer Science, The Weizmann Institute, Rehovot 76100, Israel. E-mail: p...
Lecture Notes in Computer Science, 2007
This paper focuses on local computations of distributed aggregation problems on fixed graphs. We ... more This paper focuses on local computations of distributed aggregation problems on fixed graphs. We define a new metric on problem instances, Veracity Radius (VR), which captures the inherent possibility to compute them locally. We prove that VR yields a tight lower bound on output-stabilization time, i.e., the time until all nodes fix their outputs, as well as a lower bound on quiescence time. We present an efficient aggregation algorithm, I-LEAG, which reaches both output stabilization and quiescence within a time that is proportional to the VR of the problem instance, and is also efficient in terms of per-node communication and memory. We empirically show that the VR metric also effectively captures the performance of previously suggested efficient aggregation protocols, and that I-LEAG significantly outperforms these protocols in several respects. Categories and Subject Descriptors C.4 [Performance of Systems]: [Performance attributes];
Proceedings of the 5th Jerusalem Conference on Information Technology, 1990. 'Next Decade in Information Technology', 1990
ABSTRACT
International Journal of Parallel Programming, 2006
Case-based reasoning (CBR) systems rely on the conceptual ordering of entities called cases. If a... more Case-based reasoning (CBR) systems rely on the conceptual ordering of entities called cases. If atomic case features are allowed to assume numeric as well as symbolic values, then a systematic comparison regime is needed to aggregate similarity scores. A common approach to deal with real-numbered features is normalisation. However, there are two conspicuous problems with this procedure: the similarity between two features is dependent on the corresponding values of all other cases to be ranked; and real-numbered features are often interpreted by human experts according to conceptual constraints associated with features. In such situations, a conceptual distance between two features should be determined rather than the length of a `gap' on a linear scale. Within the framework of a comprehensive case-knowledge architecture, the notion of a concept frame that can be associated with a case feature is proposed. Through this component it is possible to represent polymorphic atomic cas...
Proceedings Second International Workshop on High-Level Parallel Programming Models and Supportive Environments, 1997
Many attempts are made to integrate the resources and services of distributed computational envir... more Many attempts are made to integrate the resources and services of distributed computational environments into vir-tual parallel machines (VPMs). While utilizing idle CPU cycles and thus being very cheap and available to everyone, such metacomputing environments will ...
Lecture Notes in Computer Science, 2007