scispace - formally typeset
Search or ask a question
Author

Alexander Kesselman

Bio: Alexander Kesselman is an academic researcher from Google. The author has contributed to research in topics: Competitive analysis & Network packet. The author has an hindex of 29, co-authored 82 publications receiving 2804 citations. Previous affiliations of Alexander Kesselman include Ben-Gurion University of the Negev & Tel Aviv University.


Papers
More filters
Journal ArticleDOI
TL;DR: It is proved that the greedy algorithm that drops the earliest packets among all low-value packets is the best greedy algorithm, and the competitive ratio of any on-line algorithm for a uniform bounded-delay buffer is bounded away from 1, independent of the delay size.
Abstract: We consider two types of buffering policies that are used in network switches supporting Quality of Service (QoS). In the FIFO type, packets must be transmitted in the order in which they arrive; the constraint in this case is the limited buffer space. In the bounded-delay type, each packet has a maximum delay time by which it must be transmitted, or otherwise it is lost. We study the case of overloads resulting in packet loss. In our model, each packet has an intrinsic value, and the goal is to maximize the total value of transmitted packets. Our main contribution is a thorough investigation of some natural greedy algorithms in various models. For the FIFO model we prove tight bounds on the competitive ratio of the greedy algorithm that discards packets with the lowest value when an overflow occurs. We also prove that the greedy algorithm that drops the earliest packets among all low-value packets is the best greedy algorithm. This algorithm can be as much as 1.5 times better than the tail-drop greedy policy, which drops the latest lowest-value packets. In the bounded-delay model we show that the competitive ratio of any on-line algorithm for a uniform bounded-delay buffer is bounded away from 1, independent of the delay size. We analyze the greedy algorithm in the general case and in three special cases: delay bound 2, link bandwidth 1, and only two possible packet values. Finally, we consider the off-line scenario. We give efficient optimal algorithms and study the relation between the bounded-delay and FIFO models in this case.

194 citations

Patent
22 Dec 2009
TL;DR: In this article, a scan of the index is performed by a device of a group of devices in a distributed data replication system, the index being replicated while the objects are stored locally by the plurality of devices.
Abstract: A method is performed by a device of a group of devices in a distributed data replication system. The method includes storing an index of objects in the distributed data replication system, the index being replicated while the objects are stored locally by the plurality of devices in the distributed data replication system. The method also includes conducting a scan of at least a portion of the index and identifying a redundant replica(s) of the at least one of the objects based on the scan of the index. The method further includes de-duplicating the redundant replica(s), and updating the index to reflect the status of the redundant replica.

187 citations

Book ChapterDOI
30 Jun 2003
TL;DR: The number of steps required to reach a pure Nash Equilibrium in a load balancing scenario where each job behaves selfishly and attempts to migrate to a machine which will minimize its cost is studied.
Abstract: We study the number of steps required to reach a pure Nash Equilibrium in a load balancing scenario where each job behaves selfishly and attempts to migrate to a machine which will minimize its cost. We consider a variety of load balancing models, including identical, restricted, related and unrelated machines. Our results have a crucial dependence on the weights assigned to jobs. We consider arbitrary weights, integer weights, K distinct weights and identical (unit) weights. We look both at an arbitrary schedule (where the only restriction is that a job migrates to a machine which lowers its cost) and specific efficient schedulers (such as allowing the largest weight job to move first).

183 citations

Patent
08 Feb 2011
TL;DR: A distributed storage system has multiple instances, and at least some of the local instances are at physically distinct geographic locations as discussed by the authors, where each local instance stores metadata for the respective set of blobs in a metadata store distinct from the data stores.
Abstract: A distributed storage system has multiple instances. There is a plurality of local instances, and at least some of the local instances are at physically distinct geographic locations. Each local instance is configured to store data for a non-empty set of blobs in a plurality of data stores having a plurality of distinct data store types. In addition, each local instance stores metadata for the respective set of blobs in a metadata store distinct from the data stores. There is also a plurality of global instances. Each global instance is configured to store data for zero or more blobs in zero or more data stores and store metadata for all blobs stored at any local or global instance. The system selects one global instance to run a replication module that replicates blobs between instances according to blob policies. Some systems also include dynamic replication based on user needs.

160 citations

Journal ArticleDOI
TL;DR: In this article, the authors present a very simple distributed algorithm for computing a small CDS with an approximation factor of at most 6.91, improving upon the previous best-known approximation of 8 due to Wan et al. [2002].
Abstract: Several routing schemes in ad hoc networks first establish a virtual backbone and then route messages via backbone nodes. One common way of constructing such a backbone is based on the construction of a connected dominating set (CDS). In this article we present a very simple distributed algorithm for computing a small CDS. Our algorithm has an approximation factor of at most 6.91, improving upon the previous best-known approximation factor of 8 due to Wan et al. [2002]. The improvement relies on a refined analysis of the relationship between the size of a maximal independent set and a minimum CDS in a unit disk graph. This subresult also implies improved approximation factors for many existing algorithm.

157 citations


Cited by
More filters
Patent
11 Jan 2011
TL;DR: In this article, an intelligent automated assistant system engages with the user in an integrated, conversational manner using natural language dialog, and invokes external services when appropriate to obtain information or perform various actions.
Abstract: An intelligent automated assistant system engages with the user in an integrated, conversational manner using natural language dialog, and invokes external services when appropriate to obtain information or perform various actions. The system can be implemented using any of a number of different platforms, such as the web, email, smartphone, and the like, or any combination thereof. In one embodiment, the system is based on sets of interrelated domains and tasks, and employs additional functionally powered by external services with which the system can interact.

1,462 citations

Journal ArticleDOI
TL;DR: The Generalized Nash Equilibrium Problem is an important model that has its roots in the economic sciences but is being fruitfully used in many different fields and its main properties and solution algorithms are discussed.
Abstract: The Generalized Nash Equilibrium Problem is an important model that has its roots in the economic sciences but is being fruitfully used in many different fields. In this survey paper we aim at discussing its main properties and solution algorithms, pointing out what could be useful topics for future research in the field.

838 citations

Journal ArticleDOI
TL;DR: This paper develops a strategy to coordinate the charging of autonomous plug-in electric vehicles (PEVs) using concepts from non-cooperative games and demonstrates that convergence to the Nash equilibrium occurs very quickly over a broad range of parameters.
Abstract: This paper develops a strategy to coordinate the charging of autonomous plug-in electric vehicles (PEVs) using concepts from non-cooperative games. The foundation of the paper is a model that assumes PEVs are cost-minimizing and weakly coupled via a common electricity price. At a Nash equilibrium, each PEV reacts optimally with respect to a commonly observed charging trajectory that is the average of all PEV strategies. This average is given by the solution of a fixed point problem in the limit of infinite population size. The ideal solution minimizes electricity generation costs by scheduling PEV demand to fill the overnight non-PEV demand “valley”. The paper's central theoretical result is a proof of the existence of a unique Nash equilibrium that almost satisfies that ideal. This result is accompanied by a decentralized computational algorithm and a proof that the algorithm converges to the Nash equilibrium in the infinite system limit. Several numerical examples are used to illustrate the performance of the solution strategy for finite populations. The examples demonstrate that convergence to the Nash equilibrium occurs very quickly over a broad range of parameters, and suggest this method could be useful in situations where frequent communication with PEVs is not possible. The method is useful in applications where fully centralized control is not possible, but where optimal or near-optimal charging patterns are essential to system operation.

807 citations

Journal Article
TL;DR: A framework that is based on learning the confidence interval around the value function or the Q-function and eliminating actions that are not optimal (with high probability) is described and a model-based and model-free variants of the elimination method are provided.
Abstract: We incorporate statistical confidence intervals in both the multi-armed bandit and the reinforcement learning problems. In the bandit problem we show that given n arms, it suffices to pull the arms a total of O((n/e2)log(1/δ)) times to find an e-optimal arm with probability of at least 1-δ. This bound matches the lower bound of Mannor and Tsitsiklis (2004) up to constants. We also devise action elimination procedures in reinforcement learning algorithms. We describe a framework that is based on learning the confidence interval around the value function or the Q-function and eliminating actions that are not optimal (with high probability). We provide a model-based and a model-free variants of the elimination method. We further derive stopping conditions guaranteeing that the learned policy is approximately optimal with high probability. Simulations demonstrate a considerable speedup and added robustness over e-greedy Q-learning.

604 citations

Patent
28 Sep 2012
TL;DR: In this article, a virtual assistant uses context information to supplement natural language or gestural input from a user, which helps to clarify the user's intent and reduce the number of candidate interpretations of user's input, and reduces the need for the user to provide excessive clarification input.
Abstract: A virtual assistant uses context information to supplement natural language or gestural input from a user. Context helps to clarify the user's intent and to reduce the number of candidate interpretations of the user's input, and reduces the need for the user to provide excessive clarification input. Context can include any available information that is usable by the assistant to supplement explicit user input to constrain an information-processing problem and/or to personalize results. Context can be used to constrain solutions during various phases of processing, including, for example, speech recognition, natural language processing, task flow processing, and dialog generation.

593 citations