A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A database benchmark for high-throughput workflow management. In Proceedingsof the 5th Int. Conferenceon Extending Database Technology (EDBT96), Avignon, France, 3 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....functionality involves finding an appropriate representation for the computation. We have chosen the notion of process, similar to that used in workflow management systems (although the final implementation is rather different since workflow tools are not entirely adequate for virtual laboratories [16, 7]) A process is an annotated directed graph where the nodes represent tasks and the arcs represent the control data flow between these tasks. Thus, the notion of process allows one to capture sequences of invocations of computer programs in a distributed and heterogeneous environment and the ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A database benchmark for high-throughput workflow management. In Proceedingsof the 5th Int. Conferenceon Extending Database Technology (EDBT96), Avignon, France, 3 1996.

.... In scienti c and engineering programs, basic operations include Fourier transforms, matrix inversion, least squares tting, and operations on DNA sequences [GRS94; MBDH83] In work ow management systems, elementary operations can include any number of application programs and legacy systems [BSR96] In all cases, the elementary operations are building blocks from which larger programs and software systems are built. Although elementary operations can vary dramatically, the logic for combining them does not. In fact, the same control features arise over and over again. These features ....

A. Bonner, A. Shru, and S. Rozen. LabFlow-1: a database benchmark for high-throughput work ow management. In Intl. Conference on Extending Database Technology, number 1057 in Lecture Notes in Computer Science, pages 463-478, Avignon, France, March 25-29 1996. Springer-Verlag.

....involves finding an appropriate representation for the computation. We have chosen the notion of process, similar to that used in workflow management systems (although the final implementation is rather different since workflow tools are not entirely adequate for virtual laboratories [MVW96, BSR96] A process is an annotated directed graph where the nodes represent tasks and the arcs represent the control data flow between these tasks. Thus, the notion of process allows one to capture sequences of invocations of computer programs in a distributed and heterogeneous environment and the ....

Bonner, A., Shrufi, A., and Rozen, S. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proceedings of the Fifth International Conference on Extending Database Technology (EDBT96), Avignon, France, March 1996.

....policies (e.g. security, authentication) audit management, process monitoring, tracking, and reporting of data generated during workflow enactment. WFMSs today, are being used to re engineer, streamline, automate, and track organizational processes [JAD 94, GHS95, Fis95, SKM 96, BSR96] There has been a growing acceptance of 1 workflow technology in numerous application domains such as telecommunications, software engineering, manufacturing, production, finance and banking, laboratory sciences, health care, shipping and office automation. The current state of the art in WFMSs ....

....thereby providing a scalable, and fault tolearant architecture (for details, see Section 4.7) 2. 2 Workflow Infrastructure In the past, database technology has been used to support process management (using triggers and stored procedures) in relatively homogeneous workflow environments [DHL90, BSR96] The need for distributed object based infrastructure technology for development of large scale enterprisewide WFMS was voiced in [GHS95] Today, there exists a variety of middleware technologies and standards to support distributed computing such as TP monitors, DCE, OpenDoc, CORBA, DCOM, ....

A. Bonner, A. Shruf, and S. Rozen. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proc. of the 5th. Intnl. Conference on Extending Database Technology, pages 25--29, Avignon, France, March 1996.

....applications. Today, PSS are used for tasks such as supporting enterprises in streamlining and automating their business processes [GHS95, LA94] supporting health care organizations in coordinating immunization campaigns [SKM 96] helping scientists perform and analyze experiments [MVW96, BSR96, AH97] and allowing software development teams organize joint work on shared resources [BK94, TKP94] With the proliferation of PSS technology, the requirements in terms of availability increase but, unfortunately, current commercial systems do not provide satisfactory solutions in this area ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A database benchmark for high throughput workflow management. In Proc. of the Fifth International Conferenceon Extending Database Technology(EDBT96), Avignon, France, March 1996.

....at any point for status information that may be defined independently by arbitrary external systems or for the workflow results and any information related to their creation is not there. Scientific workflows have been explicitly addressed mainly by two projects: WASA [12] and LabFlow1 LabBase [2]. But again, the goals and approach of these efforts have been different from ours. WASA uses a commercial WFMS on top of a DBMS extended by advanced features, and several user interface, decision support, and analysis tools that offer a useful front end to scientific workflow management. WASA ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Proc. Fifth International Conference on Extending Database Technology, pages 463--478, Avignon, France, March 1996.

....focus on a particular type of process. The situation resembles in many ways that of databases before the adoption of the relational model. It is difficult to generalize individual solutions as proven by recent attempts to use commercial workflow products to support scientific applications [MVW96, BSR96] Indeed, a generic notion of process can be compared Towards a Platform for Distributed Application Development 11 with SQL. SQL alleviates the task of data management by providing a universal interface to data management tools (the database) A generic notion of process would alleviate the ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proceedings of the Fifth International Conference on Extending Database Technology (EDBT96), Avignon, France, March 1996.

....in many areas, in particular in those where computation is based on cluster of PC s and workstations or on environments involving heterogeneous platforms and applications. A careful analysis of areas such as business environments [Fry94] software engineering [BK94] or scientific data management [ILGP96, BSR96] reveals a surprising number of problems that are both pervasive and common to all of them. Such pervasiveness may explain the attention being devoted to workflow products, which are the current process support systems. It may also explain why some researchers consider workflow management to be ....

....notion of process used in this paper are workflow management systems. Existing workflow tools, however, target al..most exclusively either business processes or imaging systems, suffer from severe limitations related to performance and functionality [AAEM97] and are not easy to apply in other areas [MVW96, BSR96]. Workflow management, however, is a first step towards supporting the development of complex applications over distributed systems using already existing tools. Its concepts can be generalized by extending the notion of process to any arbitrary sequence of tool invocations (a script or pipelined ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proceedings of the Fifth International Conference on Extending Database Technology (EDBT96), Avignon, France, March 1996.

....over the database storing the states of the processes. Surprisingly, these problems are pervasive and appear in many different application areas, from virtual enterprises and business environments [Fry94, LA94] to software engineering [CKO92, TKP94, BDMQ95] and scientific data management [ILGP96, BSR96] Such pervasiveness explains the success of workflow management concepts and products, which are the most recent incarnation of process support systems. Existing workflow tools, however, are often tailored to a particular domain. In particular, existing systems target in most cases either ....

.... [AAEM97] with a few research prototypes addressing other areas [BK94, ILGP96] Such narrow purpose design is combined with severe limitations related to performance and functionality [AS96, AAEM97] limitations that further restrict the applicability of these systems (see for example [MVW96, BSR96] reporting on attempts to use commercial workflow products to support scientific applications) The project OPERA (Open Process Engine for Reliable Activities) currently ongoing at the Database Research Group of ETH Zurich, has as its main objective to paliate such glaring limitations of the ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proceedings of the Fifth International Conference on Extending Database Technology (EDBT96), Avignon, France, March 1996.

.... as any down time may not be acceptable and scalability (to support the high workloads) Final application for our discussion in this area involves the use of workflows in a high throughput mission critical application system of tracking experimental data at the Center for Genome Research [BSR96] Workflow automation is used in the laboratory information systems setting to automate the handling of samples, testing, instrumentation, data capture, and 6 Amit Sheth and Krys J. Kochut tracking of event histories. The DBMS is primarily used in this project to control and track sample ....

....J. Kochut tracking of event histories. The DBMS is primarily used in this project to control and track sample collection workflows. The throughput of such experiments range in the order of approximately 15,000 transactions per day, with peak rates reaching 22.5 queries and updates per second [BSR96] To be effective in such high throughput production environments, a lot is desired in terms of scalability, efficiency, and reliability of the underlying WFMS infrastructure and the processing entities or resources that perform the high frequency tasks. The authors address the requirements of ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shruf, and S. Rozen. LabFlow-1: A Database Benchmark for High Throughput Workflow Management. In Proc. of the 5th. Intl. Conference on Extending Database Technology, pages 25--29, Avignon, France, March 1996.

.... In scientific and engineering programs, basic operations include Fourier transforms, matrix inversion, least squares fitting, and operations on DNA sequences [GRS94; MBDH83] In workflow management systems, elementary operations can include any number of application programs and legacy systems [BSR96] In all cases, the elementary operations are building blocks from which larger programs and software systems are built. Although elementary operations can vary dramatically, the logic for combining them does not. In fact, the same control features arise over and over again. These features ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Intl. Conference on Extending Database Technology, number 1057 in Lec- REFERENCES 47 ture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

....laboratory notebook: it records what was done, when it was done, who did it, and what the results were. It can also be used to analyze workflows, to find rate limiting steps or to investigate anomalous results. This paper discusses the requirements of this DBMS. The discussion centers on LabFlow 1 [3, 2], a recently developed database benchmark for highthroughput workflow management systems (WFMSs) i:e: systems for managing high volume, missioncritical workflows. LabFlow 1 is based on the data and workflow management needs of a large genome laboratory, and reflects their real world experience. ....

....highthroughput workflow management systems (WFMSs) i:e: systems for managing high volume, missioncritical workflows. LabFlow 1 is based on the data and workflow management needs of a large genome laboratory, and reflects their real world experience. An overview of the benchmark can be found in [3], and a detailed description in [2] Benchmark software is available at the following web site: ftp: db.toronto.edu pub bonner papers workflow software Although it is based on genome laboratory workflow, we believe that LabFlow 1 captures the database requirements of a common class workflow ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Proceedings of the International Conference on Extending Database Technology (EDBT), number 1057 in Lecture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

....laboratory notebook: it records what was done, when it was done, who did it, and what the results were. It can also be used to analyze workflows, to find rate limiting steps or to investigate anomalous results. This paper discusses the requirements of this DBMS. The discussion centers on LabFlow 1 [3, 2], a recently developed database benchmark for highthroughput workflow management systems (WFMSs) i:e: systems for managing high volume, missioncritical workflows. LabFlow 1 is based on the data and workflow management needs of a large genome laboratory, and reflects their real world experience. ....

....systems (WFMSs) i:e: systems for managing high volume, missioncritical workflows. LabFlow 1 is based on the data and workflow management needs of a large genome laboratory, and reflects their real world experience. An overview of the benchmark can be found in [3] and a detailed description in [2]. Benchmark software is available at the following web site: ftp: db.toronto.edu pub bonner papers workflow software Although it is based on genome laboratory workflow, we believe that LabFlow 1 captures the database requirements of a common class workflow management applications: those that ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. Technical report, Department of Computer Science, University of Toronto, 1995. 53 pages. Available at http://www.db.toronto.edu:8020/people/bonner/bonner.html.

....updates on a set of relations, or it could insert or delete a logical formula from a set of formulas. In some applications, such as workflow management, a state may be a combination of files and databases, and elementary operations may include any number of application programs and legacy systems [BSR96,DKRR98] In all cases, however, the elementary operations are building blocks from which larger programs and software systems are built. T R provides the logical foundations and a logic programming language for building such systems. To achieve the needed flexibility, T R treats a database as a ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Intl. Conference on Extending Database Technology, number 1057 in Lecture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

....and from the U.S. Department of Energy under contract DE FG02 95ER62101. these two systems and highlighted the critical importance of being able to control locality of reference to persistent data. Details of the benchmark and the test results can be found in [3] and to a lesser extent in [4]. 1.1 Workflow Management Examples of workflow management are found in a wide range of industries, from banking and insurance, to telecommunications and manufacturing, to pharmaceuticals and health care [13, 33, 24, 11] The task is to coordinate the various activities involved in running an ....

....on a variety of systems. Although LabFlow 1 is intended to be a general benchmark for DBMSs, we have so far used it to compare storage managers only. This is achieved by running the benchmark on versions of LabBase implemented on top of different object storage managers, as described above. In [3, 4], we compare ObjectStore (version 3.0) 20, 27] and Texas (version 0.3) 34, 37] Compared to relational systems, these storage managers have been used in few production applications, so this analysis is interesting in its own right. Since they are a relatively novel technology, we compare these ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Proceedings of the International Conference on Extending Database Technology (EDBT), number 1057 in Lecture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

....Research, grant number P50 HG00098, and from the U.S. Department of Energy under contract DE FG02 95ER62101. these two systems and highlighted the critical importance of being able to control locality of reference to persistent data. Details of the benchmark and the test results can be found in [3], and to a lesser extent in [4] 1.1 Workflow Management Examples of workflow management are found in a wide range of industries, from banking and insurance, to telecommunications and manufacturing, to pharmaceuticals and health care [13, 33, 24, 11] The task is to coordinate the various ....

....object storage managers can be plugged into the DBMS. In this way, we can test a wide range of 1 In particular, the history relation never appears in the from clause of an SQL statement. existing storage managers. Our implementation of the LabFlow 1 benchmark is based on this idea [3]. We emphasize, however, that LabFlow 1 does not depend on LabBase, which is an implementation detail. In the future, we hope to use our benchmark to test the support for workflow management in off the shelf DBMSs. 3 The LabFlow 1 Benchmark The LabFlow 1 benchmark is based on the data and ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. Technical report, Department of Computer Science, University of Toronto, 1995. 53 pages. Available at http://www.db.toronto.edu:8020/people/bonner/bonner.html.

....users or multiple concurrent processes. Examples include CAD, office automation, collaborative work, manufacturing control, and workflow management. Such applications combine database transactions, application programs, and other activities into larger information systems and business processes [32, 40, 39, 10, 42, 34, 21, 43, 45]. These applications require new transaction models, new methods of transaction management, and new transaction languages [32, 40, 39, 29] This paper focuses on the latter issue: languages for specifying non classical transactions, and for combining them into complex processes. In particular, we ....

....to arbitrary graphs without much difficulty. The workflow model in the example is based on workflows at the Whitehead Institute MIT Center for Genome Research, in which workflows are used to carry out the tens of millions of biological experiments needed to map and sequence the human genome [10, 67]. Example 4.2 (Workflow) Consider a workflow consisting of a sequence of three activities. The workflow processes a stream of work items. Each item is processed first by Activity 1, then by Activity 2, and finally by Activity 3. Although the activities appear to be sequential, they execute ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for highthroughput workflow management. In Intl. Conference on Extending Database Technology, number 1057 in Lecture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

....high volume, mission critical workflows; frequent workflow change and process re engineering; an audit trail of workflow activity; and complex structured data. Due to space limitations, this paper can provide only an informal description of the benchmark. A detailed description can be found in [2]. Benchmark software is available at the following web site: ftp: db.toronto.edu pub bonner papers workflow software 2 Workflow in LabFlow 1 As noted above, an important component of many workflow management systems is a DBMS for tracking workflow. This DBMS maintains an audit trail, or event ....

....(though similar) versions of a step. Usually, the most recent version is of greatest interest to scientists, since it represents the most up to date results. LabBase uses special purpose storage and access structures to rapidly retrieve most recent results from the event history of each material [2]. Notice that as described above, the database schema depends on the workflow. For each kind of workflow step, there is a class in the database, and for each measurement made by the step, the class has an attribute. Consequently, workflow changes are reflected in the database as schema changes. In ....

[Article contains additional citation context not shown here]

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. Technical report, Department of Computer Science, University of Toronto, 1995. 53 pages. Available at http:// www.db.toronto.edu:8020/people/bonner/bonner.html.

....blocks. In scientific and engineering programs, basic operations include Fourier transforms, matrix inversion, least squares fitting, and operations on DNA sequences [14, 20] In workflow management systems, elementary operations can include any number of application programs and legacy systems [4]. In all cases, the elementary operations are building blocks from which larger programs and software systems are built. Although elementary operations vary dramatically, the logic for combining them does not. In fact, the same control features arise over and over again. These features include ....

A. Bonner, A. Shrufi, and S. Rozen. LabFlow-1: a database benchmark for high-throughput workflow management. In Intl. Conference on Extending Database Technology, number 1057 in Lecture Notes in Computer Science, pages 463--478, Avignon, France, March 25--29 1996. Springer-Verlag.

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