[Paper Review] R-GMA: First results after deployment

Rob Byrom, Brian Coghlan+19 more|ArXiv.org|May 30, 2003

Distributed and Parallel Computing Systems1 references10 citations

TL;DR

R-GMA is a relational Grid monitoring architecture that leverages SQL-based data modeling to provide a unified, queryable view of distributed grid information across Virtual Organizations. It supports multiple query types (history, latest, continuous) and five distinct Producer types, with a mediator routing queries to optimal Producers; deployment shows support for 40–50 sites publishing every 30 seconds, with performance bottlenecks identified for optimization.

ABSTRACT

We describe R-GMA (Relational Grid Monitoring Architecture) which is being developed within the European DataGrid Project as an Grid Information and Monitoring System. Is is based on the GMA from GGF, which is a simple Consumer-Producer model. The special strength of this implementation comes from the power of the relational model. We offer a global view of the information as if each VO had one large relational database. We provide a number of different Producer types with different characteristics; for example some support streaming of information. We also provide combined Consumer/Producers, which are able to combine information and republish it. At the heart of the system is the mediator, which for any query is able to find and connect to the best Producers to do the job. We are able to invoke MDS info-provider scripts and publish the resulting information via R-GMA in addition to having some of our own sensors. APIs are available which allow the user to deploy monitoring and information services for any application that may be needed in the future. We have used it both for information about the grid (primarily to find what services are available at any one time) and for application monitoring. R-GMA has been deployed in Grid testbeds, we describe the results and experiences of this deployment.

Motivation & Objective

• To design a scalable, interoperable Grid Information and Monitoring System based on the GGF GMA model.
• To leverage the expressive power of the relational model for unified, complex querying across distributed grid data sources.
• To support diverse monitoring workloads through specialized Producer types (e.g., streaming, resilient, latest).
• To enable efficient, dynamic discovery and routing of data via a mediator that selects optimal Producers for each query.
• To deploy and evaluate the system in real Grid testbeds to assess performance and robustness.

Proposed method

• R-GMA implements the GMA Consumer-Producer model using relational databases, where Producers register via SQL CREATE TABLE and publish using SQL INSERT.
• It supports five Producer types: DataBaseProducer (persistent RDBMS storage), StreamProducer (in-memory streaming), ResilientStreamProducer (disk-backed streaming), LatestProducer (latest-value storage), and CanonicalProducer (code-triggered queries).
• Consumers issue SQL SELECT queries, which the mediator routes to the most appropriate Producers based on query type and data availability.
• The system uses a Registry (directory service) to register Producers and Consumers, with soft-state registration for fault tolerance.
• It is built on Java servlets and supports multiple APIs (Java, C++, C, Python, Perl) via SWIG, with plans to migrate to OGSA web services.
• The mediator dynamically discovers and connects to the best Producers for each query, enabling distributed, scalable data retrieval.

Experimental results

Research questions

• RQ1 Can a relational model effectively unify and query heterogeneous grid monitoring data across distributed, autonomous sources?
• RQ2 How does the performance of a relational Grid monitoring system scale under concurrent publishing from multiple sites?
• RQ3 What are the performance bottlenecks in a high-throughput, low-latency Grid monitoring system using SQL-based Producers?
• RQ4 Can a mediator-based routing system efficiently select optimal Producers for different query types (history, latest, continuous)?
• RQ5 How effective is soft-state registration in maintaining system resilience during transient failures?

Key findings

• R-GMA successfully supports the deployment of 40 to 50 grid sites publishing monitoring data every 30 seconds, demonstrating scalability in a real testbed.
• Buffering tuples in producers had little effect on system throughput, indicating that performance bottlenecks lie elsewhere, such as in Tomcat or JVM configuration.
• The system achieved stable operation with a mix of Producer types, including an Archiver feeding into a LatestProducer, confirming data consistency and flow integrity.
• Performance tuning is expected to improve throughput through adjustments to Tomcat settings, JVM parameters, and increased physical memory.
• A future architectural shift to distributed query execution across multiple LatestProducer archives could significantly enhance scalability and reduce load on individual nodes.
• The system’s reliance on soft-state registration proved effective for maintaining up-to-date discovery of Producers and Consumers, ensuring robustness.