Load Shedding in a Data Stream Manager

Tatbul, Nesime; Çetintemel, Uǧur; Zdonik, Stanley B.; Cherniack, Mitch; Stonebraker, Michael

doi:10.1016/b978-012722442-8/50035-5

Cited by 407 publications

(334 citation statements)

References 1 publication

Supporting

Mentioning

334

Contrasting

Order By: Relevance

“…Stream systems such as Aurora [13] use semantic shedding [38] as one of the techniques to decide which tuples to drop when resources run low-that is, the less useful the data is for the recipient, the earlier it gets dropped. Here we take this concept further by deciding to drop (or generalize) tuples when a user's privacy would be violated.…”

Section: Related Workmentioning

confidence: 99%

Middleware for Pervasive Spaces: Balancing Privacy and Utility

et al. 2009

View full text Add to dashboard Cite

Abstract. Middleware for pervasive spaces has to meet conflicting requirements. It has to both maximize the utility of the information exposed and ensure that this information does not violate users' privacy. In order to resolve these conflicts, we propose a framework grounded in utility theory where users dynamically control the level of disclosure about their information. We begin by providing appropriate definitions of privacy and utility for the type of applications that would support collaborative work in an office environment-current definitions of privacy and anonymity do not apply in this context. We propose a distributed solution that, given a user's background knowledge, maximizes the utility of the information being disclosed to information recipients while meeting the privacy requirements of users. We implement our solution in the context of a real pervasive space middleware and provide experiments that demonstrate its behaviour.

show abstract

Section: Related Workmentioning

confidence: 99%

Middleware for Pervasive Spaces: Balancing Privacy and Utility

et al. 2009

View full text Add to dashboard Cite

show abstract

“…This is also referred to as load shedding. Most current approaches are based on random load shedding [6,33,42] (Tatbul et al [42] also consider simple heuristics for semantic load shedding, where certain events have higher value to the query than others). Random load shedding works well for queries that compute aggregates, but as we show [17], random load shedding is inferior if we are concerned with the approximation quality of set-valued query results.…”

Section: Techniques For the Load Smoothing Modulementioning

confidence: 99%

The Architecture of the Cornell Knowledge Broker

Demers

Gehrke

Riedewald

2004

Intelligence and Security Informatics

View full text Add to dashboard Cite

Abstract.Intelligence applications have to process massive amounts of data in order to extract relevant information. This includes archived historical data as well as continuously arriving new data. We propose a novel architecture that addresses this problem -the Cornell Knowledge Broker. It will not only support knowledge discovery, but also security, privacy, information exchange, and collaboration.

show abstract

“…Thus it routes data through operators adaptively, based on arrival characteristics. Aurora [33,34] focuses on QoS-and memory-aware operator scheduling and load shedding for coping with transient spikes in data.…”

Section: Adaptive Query Processing With Relational Datamentioning

confidence: 99%

FMware: Middleware for Efficient Filtering and Matching of XML Messages with Local Data

Candan¹,

Dönderler²,

Yan³

et al. 2006

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. XML message filtering systems are used for sifting through real-time messages to support business data mining and reporting. An XML message filtering system needs to (a) process registered filter predicates on multiple distributed real-time streams and (b) match and validate the filter results with local data to identify the relevant data that can be used for higher-level processing. Although efficient real-time filtering schemes exists, the matching phase of the operation where filter results have to be matched against local data to select those matches that are relevant to the particular task remains to be expensive as it requires expensive join operations. In this paper, we present an efficient middleware (FMware) for filtering and matching XML messages against locally available data. The proposed operator relies on a novel clusterdomain matching scheme to reduce the cost of the process. We analytically study the cost of the proposed middleware and experimentally show that it adaptively reduces the number of local data accesses and provides large savings in matching time with respect to cluster-unaware matching.

show abstract

Load Shedding in a Data Stream Manager

Cited by 407 publications

References 1 publication

Middleware for Pervasive Spaces: Balancing Privacy and Utility

Middleware for Pervasive Spaces: Balancing Privacy and Utility

The Architecture of the Cornell Knowledge Broker

FMware: Middleware for Efficient Filtering and Matching of XML Messages with Local Data

Contact Info

Product

Resources

About