C. Isert and K. Schwan. ACDS: Adapting Computational Data Streams for High Performance. In 14th International Parallel and Distributed Processing Symposium (IPDPS'00), April 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....such as OpenORB [3] and Bossa Nova [14] o#er a flexible infrastructure that supports QoSaware composition and reflection. While these frameworks do not provide specific streaming support, they can serve as a basis for building information flow middleware. Event based middleware such as Echo [7, 11] provides a type safe and e# cient way of communicating data and control information in a distributed and heterogeneous environment. A higher level Infopipe layer can also be built on top of these platforms. Ensemble [34] and DaCaPo [27] are protocol frameworks that support the composition and ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance computing. In International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....index file for better performance. R trees are used as the indexing method for summary and detailed index files. 2) Processing of Data: Filters and Streams: Recent research on programming models for developing applications in the Grid has converged on the use of component based models [1] 20] [25], 31] 32] in which an application is composed of multiple interacting computational objects. In the DataCutter project, we have developed a framework, called filter stream programming, for developing data intensive applications in a distributed environment. The filter stream programming model ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000.

....issues for data analysis applications. This paper differs from our previous work in that we look at the effect on data reuse of functional decomposition. There are a number of research projects that focus on component based models for developing applications in a distributed environment [1, 3, 13, 15, 24, 31, 32, 33, 38]. In these models, the application processing structure is decomposed into a set of interacting computation components. The earlier work on component based frameworks has focused on improving the performance of a single or a set of independent queries by effectively decomposing the application ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000.

....data sources to the client. In this work we address the scheduling of multiple queries, represented as pipelined chain of operations on data, in a Grid environment. Our approach focuses on component based frameworks, where an application is developed from a set of interacting software components [13, 18, 20, 28, 27]. In a component based framework, the placement of components onto computational resources represents an important degree of flexibility in optimizing application performance. Network and computation overheads can be decreased by efficiently placing components to deal with computational ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000.

....storage systems) offers a powerful and flexible environment for developing and deploying applications that analyze large datasets. Component based frameworks and services have been gaining acceptance as a viable approach for application development and execution in distributed environments [1, 4, 11, 13, 14, 18, 23, 24, 27, 29]. Such models facilitate the implementation of applica p1 p2 p3 (a) b) c) Figure 1. An application server may use many different parallel configurations depending on what is most efficient for an application. a) shared memory, b) distributed shared memory, or (c) distributed ....

.... (GrADS) and support mechanisms for storing metainformation needed to control program execution the Grid Information System (GIS) Several research projects have investigated the design, implementation, and application of component based frameworks for application development and deployment [1, 11, 13, 14, 24, 27, 29]. The CommonComponent Architecture Project (CCA) 17] leads a standardization effort for building distributed software component systems for scientific and engineering applications. The Open Grid Services Architectures (OGSA) effort [21] draws from concepts and technologies that evolved ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of the 2000.

....[59] and remote I O [57] However, providing support for efficient subsetting and processing of very large scientific datasets stored in archival storage systems in a distributed environment remains a challenging research issue. Component based programming models are becoming widely accepted [19,33,38,50,53] for developing applications in distributed, heterogeneous environments. In this model, the processing structure of an application is represented as multiple objects that interact with each other by moving data and control information. We have developed a component based framework, called ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS

....progresses from the data source(s) to the client, and on the ability to move all or part of its computations to other machines that are well suited for the computation. Recent research on programming models for developing applications in the Grid has converged on the use of component based models [2, 4, 5, 6, 7, 9, 10], in which an application is composed of multiple interacting computational objects. In the DataCutter project [4, 5, 8] we are developing a framework, called filter stream programming, for developing data intensive applications in a distributed environment. This model represents components of a ....

....with assigning work to existing filter instances. This approach is essential for any practical Grid framework that supports adaptivity. 4. Related Work There is an increasing number of research projects that focus on component based models for developing applications in a distributed enviroment [2, 9, 6, 7, 10]. The ABACUS framework [3] addresses the automatic and dynamic placement of functions in data intensive applications between clients and storage servers. This work is closely related to DataCutter in that application components 6 Total Scratch Memory Allocated 0mb 64mb 128mb 192mb 256mb 320mb ....

[Article contains additional citation context not shown here]

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000), pages 641--646, Cancun, Mexico, May 2000. IEEE Computer Society Press.

....are implemented as C classes and are statically linked to ABACUS system on server and client nodes, i.e. object migration is carried out by moving the state buffer and instantiating the object on the new node, rather than moving the object codes. 5. 2 The dQUOB and ACDS Systems The ACDS [18] (Adapting Computational Data Streams) from Georgia Institute of Technology is a framework that addresses construction and adaptation of computational data streams in an application. Computational data streams model application processing as computational objects associated with data streams. A ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000), pages 641--646, Cancun, Mexico, May 2000. IEEE Computer Society Press.

No context found.

C. Isert and K. Schwan. ACDS: Adapting Computational Data Streams for High Performance. In Proceedings of IPDPS, May 2000.

No context found.

Carsten Isert and Karsten Schwan, "Acds: Adapting computational data streams for high performance," in 2000.

.... to their remote maintenance, using on board communication and computer equipment [16] Finally, server systems or the overlay networks that amplify individual servers are becoming increasingly conscious of client needs, currently addressing specific tasks like remote graphics and visualization [9, 31], but with future work already considering other assistance, such as the functionality currently provided by application servers that perform XML or HTTP processing for companies large scale operational information systems [15] Service morphing. The key problem addressed by our work is how to ....

.... on runtime code specialization [14] and on code generation of embedded systems [11, 21, 32] Our research is based on our previous work with distributed, real time, adaptive, and multimedia systems [24, 27, 30] with sensor based embedded applications [23] and with high performance middleware [4, 9] and applications [31] We are also leveraging prior large scale funding from the National Science Foundation and the Department of Energy that has created the basic publish subscribe middleware used in this work [4, 33] and is currently creating the new network measurement techniques [10, 22] and ....

C. Isert and K. Schwan. Acds: Adapting computational data streams for high performance. In Proceedings of the 2000.

....data element, but merely transforms it by taking an average and forwarding the average instead. As a result, resource savings are due to reductions in required network processing at the server and in the reduction of required network bandwidth, important considerations for remote data visualization[12]. By comparison, the same filter function implemented in Java requires 13.33 msecs for execution with Just In Time compilation enabled, 75.43 msecs otherwise) This indicates that Java based specialization may have such a costly run time that server extension would no longer be a win win ....

....from simple sensor data capture and analysis to the transport and analysis of image data streams. ECho has also been used in a range of high performance applications, including a Hydrology Workbench used by collaborating scientists and a similar system used for collaboration across the Internet[12]. ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....via meaningful data sets generated at runtime by the computational models being employed. This paper demonstrates ECho s high performance across heterogeneous hardware platforms, using networked machines resident at Georgia Tech. In previous work, we have used ECho in Internet wide collaborations[14], and we have demonstrated its ability to represent both the control and the data events occurring in distributed computational workbenches. Dynamic data provision and consumption ECho supports the publish subscribe model of communication. Thus new components can be introduced into an ....

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....with them. The services placed onto such attached network processors (ANPs) 1) address large data applications by filtering data as per current end user needs, an example being filtering to transmit only those data items that are actually currently viewed in a remote scientific visualization [9, 12]; 2) they address multimedia applications, by executing scheduling algorithms that reduce data by discarding packets based on loss tolerance and deadline information [11, 17] and (3) they concern execution of codes for large scale operational information systems (OISs) used by companies that ....

C. Isert and K. Schwan. ACDS: Adapting Computational Data Streams for High Performance. In Proc. of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....in user or kernel space) 1) to alter their inter machine data communications or (2) to dynamically change the functionality of the associated Qchannel based resource management services. Useful functions include those that (i) filter user level communications based on current end user needs [5], or that (ii) compute the utility [20] or payoff [8] of certain communications, or actions, or changes to their qualities from an application perspective (thereby providing to kernel level resource managers application level interpretations of quality [21] or that (iii) implement event ....

C. Isert and K. Schwan. ACDS: Adapting Computational Data Streams for High Performance. In Intl. Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....over a variable sized window, for instance. Other examples are functions that lter messages or that compute the actual utility applications derive from receiving them[19] by recognizing and computing statistics derived from the application level header information contained in such messages[17]. Software Architecture. A sample use of dproc monitoring is one in which the current availability of cache space is noted for some set of cluster nodes used by an application program. Information about cache space is captured in the kernels of all server nodes and transported to a single dproc ....

Carsten Isert and Karsten Schwan, "ACDS: Adapting Computational Data Streams for High Performance", International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

.... D Channel E Channel F Clustering Channel B Channel A global transport Model Circulation Residual Legends: Computation Components : Access Stations Channel A ECho Event Channels cally control which data they transform and display, by controlling what is being sent to them by data sources[10]. Thus, event receivers must be able to customize event producers. JECho handles the dynamic, receiver initiated specialization of data producers with a novel software abstraction: eager handlers. An eager handler is an event handler that consists of two parts, with one part remaining in the ....

....by enabling changes both at the data consumer and provider sides of a communication, thereby reducing bandwidth needs and the processing power requirements at the recipients. We have already demonstrated the importance and benefits of client controlled, dynamic data filtering for wide area systems[10]. Such filtering is even more important in the Java environment where communication costs are high. Therefore, our principal goal in creating the notion of eager handlers is to prevent networks with limited bandwidth and event consumer stations with limited computing capabilities from being ....

C. Isert and K. Schwan, "ACDS: Adapting Computational Data Streams for High Performance, Proceedings of IPDPS '00, 2000.

.... mirroring differs from message broadcast or multicast in that it is performed at the application level. This enables us to substantially reduce mirroring traffic compared to implementations described previously [11] by filtering events based on their data types [12] and or their data contents [13], by coalescing certain events particularly when the effects of a later event overwrites those of previous events, or by simply varying mirroring rates according to current application needs concerning the consistency of mirrored vs. original data. In Section 4.2, we experimentally evaluate the ....

C. Isert and K. Schwan, "ACDS: Adapting Computational Data Streams for High Performance", In Proc. of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....with a novel approach to selectively extracting data from data streams. Earlier work done by our group has established the benefits of encapsulating needs mismatch style computations into logical tasks that can be associated with data streams to maximize proximity or availability of resources [25, 5, 17, 21]. This paper extends these notions by providing the user with an intuitive relational model for thinking about needs mismatch computations and a prototype system for creating these computations and embedding them into a data stream. Application of the work first to safety critical systems[24] ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....via meaningful data sets generated at runtime by the computational models being employed. This paper demonstrates ECho s high performance across heterogeneous hardware platforms, using networked machines resident at Georgia Tech. In previous work, we have used ECho in Internet wide collaborations [17], and we have demonstrated its ability to represent both the control and the data events occurring in distributed computational workbenches. 2 G. Eisenhauer, et al. Event Services in High Performance Systems Dynamic data provision and consumption. ECho supports the publish subscribe model of ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....difference frame is generated as Frame Diff Ev. CREATE RULE C:4 ON Frame Ev IF SELECT Frame Diff Ev FROM Frame Ev as f1, Frame Ev as f2 WHERE f1 meets f2 THEN diff 4 From Relational Schema to Event Channel Configuration At Georgia Tech our work [11] and the work of others in our group [6] has established the viability of the computational data stream. A computational data stream is a data stream with computation inserted at the source, destination, or at intermediate points between. The computations serve to transform, aggregate, or filter the data. For instance, aggregation might ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....neighbor points in a 3D space to reduce downstream bandwidth needs. Transformation might perform units conversion or partially prepare the data for visualization. Computational data streams are one of the underlying mechanisms of the Infosphere project [13] Their viability has been established in [6]. Data streams have been treated by otherse in [4] 2] and [7] Our work with dQUOB is in adapting database queries to operate over streaming data instead of database tables. Viewing data streams as data sources over which relational queries can be specified has been explored in the past in the ....

....time. Further, our work with a global atmospheric transport model and earlier with a autonomous robotics application has shown that relevant and meaningful queries can be stated with the SQL query language. Whereas earlier work by our group has justified the benefits of stream computation [6], our work introduces a conceptual model for thinking about computational data streams and demonstrates the utility of coupling computation with queries to achieve greater gains in total stream processing. The contributions of this paper are two fold. Conceptualizing streaming data with a ....

[Article contains additional citation context not shown here]

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....neighbor points in a 3D space to reduce downstream bandwidth needs. Transformation might perform units conversion or partially prepare the data for visualization. Computational data streams are one of the underlying mechanisms of the Infosphere project [16] Their viability has been established in [6] and considered by others in [3] 4] and [8] Our work with dQUOB is in adapting database queries to operate over streaming data instead of database tables. Viewing data streams as data sources over which relational queries can be specified has been explored in the past in the context of ....

....our work with a global atmospheric transport model and earlier with a autonomous robotics application has shown that queries relevant and meaningful to users accessing data streams can be stated with the SQL query language. Earlier work by our group has justified the benefits of stream computation [6], that is, performing transformations to the data as it flows from source to client. Our work has shown that by preceding a transformation computation with queries, one can significantly decrease the total amount of time spent in transformation, and additionally decrease total network bandwidth ....

[Article contains additional citation context not shown here]

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....dQUOB s ability to reduce end to end latency is demonstrated by a 99 reduction achieved by replacing a weak condition with a strong one, thereby improving the query s filtering ability. Similar results were achieved across the Internet with an ad hoc implementation of queries described in [16]. Finally, our results show that, using an application realistic action, an unoptimized query can consume up to a startling 90 of quoblet execution time, thereby demonstrating the importance of runtime reoptimization. The opportunities presented by such optimization are demonstrated in earlier ....

....reference to well defined and meaningful data attributes like longitude and latitude. In this example, filtering is performed using both (1) an explicitly defined region of interest (i.e. the arctic circle) and (2) a bounding box generated by an active user interface in response to user actions [16]. The rule, named C:1 , accepts two input event types: Data Ev, Request Ev. The IF clause delineates the query portion of the rule; the action portion is delineated by the THEN clause. The SELECT statement within the query defines the resulting event type. The result can be either a new or an ....

[Article contains additional citation context not shown here]

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....dQUOB s ability to reduce end to end latency is demonstrated by a 99 reduction achieved by replacing a weak condition with a strong one, thereby improving the query s filtering ability. Similar results were achieved across the Internet with an ad hoc implementation of queries described in [8]. Finally, our results show that, using an applicationrealistic action, an unoptimized query can consume up to a startling 90 of quoblet execution time, thereby demonstrating the importance of runtime reoptimization. The opportunities presented by such optimization are demonstrated in earlier ....

....the latter. The FROM statement defines aliases used in the query body. The query condition is nested inside the WHERE. The user desires data from upper levels (above level 30) for the region defined by the bounding box generated automatically by an active user interface in response to user actions [8]. d.timestep 12 = 0 causes all but one event per day to be discarded. The THEN clause specifies that the function ppm2ppb be executed when the query evaluates to true. Ppm2ppb converts parts permillion to parts per billion. Not shown is the function code itself which is written using a ....

[Article contains additional citation context not shown here]

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....element, but merely transforms it by taking an average and forwarding the average instead. As a result, resource savings are due to reductions in required network processing at the server and in the reduction of required network bandwidth, an important consideration for remote data visualization[12]. By comparison, the same filter function implemented in Java requires 13.33 msecs for execution with Just In Time compilation enabled, 75.43 msecs otherwise) This indicates that Java based specialization may have such a costly run time that server extension would no longer be a win win ....

....from simple sensor data capture and analysis to the transport and analysis of image data streams. ECho has also been used in a range of high performance applications, including a Hydrology Workbench used by collaborating scientists and a similar system used for collaboration across the Internet[12]. ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....via meaningful data sets generated at runtime by the computational models being employed. This paper demonstrates ECho s high performance across heterogeneous hardware platforms, using networked machines resident at Georgia Tech. In previous work, we have used ECho in Internet wide collaborations[14], and we have demonstrated its ability to represent both the control and the data events occurring in distributed computational workbenches. ffl Dynamic data provision and consumption ECho supports the publish subscribe model of communication. Thus new components can be introduced into an ....

Carsten Isert and Karsten Schwan. ACDS: Adapting computational data streams for high performance. In Proceedings of International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

....insight from our previous work is that the underlying computing network platform is typically shared by multiple end users and therefore, cannot be relied upon to provide consistent levels of service. This implies the need for runtime con guration of data streams processing actions [RSYJ97, MS98, IS00, Ise99] of the locations at which actions are performed, and of the identities or even the behavior of the network links across which data is transported [RSF97] The contribution of this paper is to demonstrate the performance bene ts derived from runtime con guration of data streams ....

Carsten Isert and Karsten Schwan. Acds: Adapting computational data streams for high performance. In International Parallel and Distributed Processing Symposium 2000, 2000. Submitted.

No context found.

C. Isert and K. Schwan. ACDS: Adapting Computational Data Streams for High Performance. In 14th International Parallel and Distributed Processing Symposium (IPDPS'00), April 2000.

No context found.

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance computing. In International Parallel and Distributed Processing Symposium (IPDPS), May 2000.

No context found.

C. Isert and K. Schwan. ACDS: Adapting computational data streams for high performance. In 14th International Parallel & Distributed Processing Symposium (IPDPS 2000.

No context found.

Isert C, Schwan K. ACDS: Adapting computational data streams for high performance computing. In International Parallel and Distributed Processing Symposium (IPDPS). 2000; .

No context found.

Isert, C. and Schwan, K. ACDS: Adapting computational data streams for high performance. 641-646. 2000.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC