Distributed and Fault Tolerant Computing

Distributed and Fault-Tolerant Computing (DFTC) addresses many of the core issues in computer science and is a critical area of research for IBM. Our researchers work at a host of research locations across the world and are interested in a wide range of topics. Our current focus is on addressing key challenges in Grid and Autonomic computing, and flexible integration of disparate IT and business processes for enabling on-demand services – all of which have a significant impact in transforming today’s businesses:

• Integration -- creating business flexibility by integration of disparate, unconnected business and IT process.
• Automation -- reducing costs and increasing business responsiveness through IT and business linkage.
• Virtualization -- improving working capital and asset utilization.

Links to a few representative projects are listed below along with a sampling of different aspects of our activities in the larger research community.

WSLA: Web Service Level Agreement for On Demand Services
WSLA is a novel framework for specifying and monitoring Service Level Agreements (SLAs) for Web Services. The framework components include tooling for online creation of SLAs, dynamic deployment of SLAs and measurement and monitoring of QoS parameters for checking the agreed-upon service levels, and reporting violations to the authorized parties involved in the SLA management process.

WSMM - Web Services Management Middleware for On Demand Services
WSMM provides fundamental support for differentiated services based on Service Level Agreements (SLAs), for web services. This enables service providers to offer the same web service, on-demand, at different performance levels (e.g., response time thresholds and throughput limits). WSMM dynamically allocates resources to web service requests, with the goal of optimizing the system utility. WSMM and WSLA technologies are available for download as part of the Emerging Technologies Toolkit.

Web Services On Demand

Rainforest
The goal of Rainforest (follow on to Océano project) is to enable policy directed dynamic componentized autonomic provisioning and management. The primary objectives are to provide a framework for heterogeneity of resources and solutions; to autonomically create complex provisioning tasks and policies from basic elements; and to utilize existing provisioning products and management systems.

ReGS- Reporting Grid Service
ReGS is an OGSA-based Meta-OS Grid Service for logging, tracing and monitoring applications in a distributed, heterogeneous computer environment, with extensible filtering capabilities. It exploits OGSI interfaces and provides standard logging interfaces for use by other Grid Services and Applications by virtualizing existing logging systems, e.g., zOS logging, NT events, and Unix syslogs.

Optimal Grid
The Optimal Grid middleware provides a grid-enabled collaboration framework, sophisticated management infrastructure, and problem-solving environment for grid computing. It is designed to help users harness the power of future grid utilities by hiding the complexities of partitioning, distributing, load balancing, and adapting a grid application to dynamic changes in available compute resources.

Gryphon
The Gryphon project focuses on advancing messaging middleware. It extends the scalable publish/subscribe framework to support efficient content-based routing in wide-area networks and stronger guarantees for message delivery by developing protocols tolerant to failures in the overlay network and in the clients.

Policy-based Computing Systems
In policy-based computing, management operations are specified in terms of the objectives or goals that need to be met, rather than the detailed instructions that need to be executed. A number of activities have been undertaken to develop policy schemas and architectures in different domains, including: automated provisioning of computing systems, auditing of configuration constraints in Storage Area Networks (SANs), and guiding analysis and problem determination processes.

XMT - Policy-based Extended Web Services and Messaging Transactions
The XMT project addresses main challenges in transactional coordination, including: defining transaction policies in an XML Web services/messaging-based environment, managing transaction policies and corresponding actions through an effective middleware system, and integrating such transaction policy system with Web services and messaging-based transaction middleware.

DSF - Data Sharing Facility
DSF is an experimental project to build a server-less file system that distributes all aspects of file and storage management over cooperating machines interconnected by a fast-switched network. DSF is aimed at scaling to hundreds of machines using commodity components. DSF provides a global memory cache, distributed file management, and distributed storage repository.

SINTRA - Distributing Trust on the Internet
SINTRA (Secure INtrusion-Tolerant Replication Architecture) is a protocol suite for secure and fault-tolerant service replication in asynchronous networks such as the Internet. Using randomization, novel customized cryptographic tools, and optimistic methods, SINTRA provides the first practical protocols that do not rely on any timing assumption, while tolerating active coordinated attacks.