ERCIM News No.45 - April 2001 [contents]
Grid-related Activities in the Laboratory of Parallel and Distributed Systems of SZTAKI
by Péter Kacsuk and Ferenc Vajda
The Laboratory of Parallel and Distributed Systems (LPDS) of SZTAKI has a long and successful history in the research and development of distributed systems, tools and applications. A number of these activities and results are closely related to the current grid-oriented projects of the laboratory. The largest computer cluster in Hungary runs under the supervision of LPDS.
In the past the laboratory developed different graphical, parallel programming environments. One of them is the P-GRADE that is an integrated set of programming tools for general message passing applications to be run in heterogeneous computing environments or supercomputers. It was developed in the frame of several EU and Hungarian projects, and supported by the Hungarian National Committee for Technological Development. The most interesting tools in the P-GRADE package are:
- GRAPNEL graphical programming language which supports parallel programming activities by graphics while the other part of a program can be written in a textural language (eg C). GRAPNEL programs can be compiled to existing message passing systems (PVM or MPI).
- GRED, a graphical editor to write parallel applications. It can be used to construct program graphs both in the Application and the Process Windows supported by GRAPNEL.
- GRM, a semi-online monitoring tool to generate and collect information (trace data or statistical information) about an application. Host and network monitoring sensors can be connected to GRM. It delivers the collected information to PROVE.
- PROVE, a visualization tool to analyze and interpret the trace file information and to present it graphically to the user during execution. GRM and PROVE provided solid bases to the current grid monitoring activities of the laboratory.
LPDS is the coordinator of a large-scale research/development project supported by a grant of the Hungarian Ministry of Education, Research and Development Division. The main goals of the project are:
- to develop a virtual supercomputer based on the computers of the institutions connected to academic computer network
- to develop and use a metacomputing supervision system to supervise the heterogeneous systems
- to support pilot-applications of the system
- as side effect of the project, to establish a knowledge center to provide professional support to the proliferation of these new technologies.
The Hungarian Scientific Research Fund (OTKA) provided a grant to the laboratory for the period of 2000-2003. The title of the project is A graphical supervisory system for geographically distributed heterogeneous metacomputing environment. The goal is to extend P-GRADE to grid applications.
Participation in the DataGrid Project
Represented by our laboratory, SZTAKI is an associate member of the DataGrid Project supported by the European Commission in the 5th frame of the IST program. The objective of the project is to enable next generation scientific exploration, which requires intensive computation and analysis of shared large-scale databases (the size from hundreds of TeraBytes to PetaBytes) across widely distributed scientific communities. It is based on emerging computational grid technologies (mainly on the Globus Project). LPDS takes part in the Monitoring Services Workpackage.
The ability to monitor and manage distributed components is critical for enabling high performance distributed computing. Monitoring data is needed to determine the source of performance problems, to support tuning systems and applications for better performance, for fault detection and recovery and for performance prediction services.
The system consists of the following main components:
- event consumers, which could be any program that requests event data for the purpose of real-time monitoring (performance analysis), overview monitoring (combining information to make some decision), process monitoring (triggering an action on an event from a server process) and archiving (providing ability to do historical analysis of system performance)
- event producers, which are gathering event data from host, network, process, application, storage, I/O, etc.,
- directory service, which locates names and describes the structural characteristics of any data available to the grid.
The main tasks of the laboratory in the project are defining and developing the producers activities and their communication principles and protocols with the consumers and event directories.
The researchers of LPDS took an active part in the European Grid Forum and now in the united international Global Grid Forum. The laboratory actively participated in the Cactus Testbed demonstration of the European Grid Forum at the Supercomputing Conference (4-10 November 2000) in Dallas. Cactus is an open source problem-solving environment designed for scientists and engineers. The demonstration simulated the collision of black holes and was executed on the machines of the six participating European institutions including the computer cluster of SZTAKI.
On the occasion of ERCIMs 10th anniversary, a Prospective Report entitled Network-based Distributed Computing (Metacomputing) by Péter Kacsuk and Ferenc Vajda was supported by an ERCIM grant. The document is available at the ERCIM website under http://www.ercim.eu/publication/prosp/.
Péter Kacsuk - SZTAKI
Tel: +36 1 329-7864
Ferenc Vajda - SZTAKI
Tel: + 36 1 329-7864