Scope

Techniques to increase the resilience and availability of stored data are fundamental to building dependable distributed systems. Past research on this topic has been very successful, leading to the emergence of new commercial products, including several replicated databases.

Yet, new demands call for further research. The need to support replication over wide-area networks, to replicate massive quantities of information in very large clusters, and to use the Internet infrastructure as the basis for building a perpetual data store, has spawned new research directions. Recent proposals, such as database replication based on group communication and peer-to-peer data storage and retrieval, illustrate the diversity of new possibilities addressed by the Workshop on Dependable Distributed Data Management.

This one-day workshop seeks contributions related to management of data in dependable distributed systems. Topics of interest include, but are not limited to, the following:

• Database replication
• Peer-to-peer data management
• Data recovery: online and disaster recovery
• Group communication vs transactional semantics
• Partial replication and fragmentation
• Scalability and performance
• Clustering and wide-area issues
• Replication middleware
• Autonomic replicated data

Registration

The workshop will take place on 17th October 2004 (one day before SRDS) at the same location as SRDS. Registration is $150 (in addition to registration fee for SRDS). Registration for the workshop is via the registration process for SRDS (found at the SRDS site).

Program

Session 1

• Welcome/About the program.
  L. Rodrigues (U. Lisboa, Portugal), and J. Pereira (U. Minho, Portugal).
• Invited paper: On the benefits of the functional modular approach to distributed data management systems.
  R. Friedman (Technion, Israel), and M. Raynal (IRISA, France).
• Understanding replication: A pratical approach using a component-based application server and a group communication system.
  M. Pasin (UFSM, Brazil), T. Weber (UFRGS, Brazil), and M. Riveill (UNSA/I3S, France).

Session 2

• Invited paper: Membership-oblivious protocols.
  F. Pedone (USI, Switzerland), P. Felber (U. Neuchatel, Switzerland), and S. Pleisch (Cornell, USA).
• An intrusion-tolerant web server based on the DISTRACT architecture.
  R. Ferraz, B. Gonçalves, J. Sequeira, M. Correia, N. Neves, and P. Veríssimo (FCUL, Portugal).
• Against attacks and faults: An autonomic approach to secure and reliable computing.
  P. Ezhilchelvan (U. Newcastle upon Tyne, UK) and R. Maxion (CMU, USA).

Session 3

• Invited paper: Efficient reliable internet storage.
  R. van Renesse (Cornell, USA).
• Invited paper: Semi-passive replication and eventual leadership.
  X. Defago (JAIST, Japan).
• Development and evaluation of database replication in ESCADA.
  A. Sousa, L.Soares, A. Correia Jr., F. Moura, R. Oliveira (U. Minho, Portugal).
• Revisiting epsilon serializability to improve the performance of the database state machine.
  A. Correia Jr., A. Sousa, L.Soares, F. Moura, R. Oliveira (U. Minho, Portugal).

Session 4

• Panel: Dependable replicated data: Strategies, drawbacks and benchmarking.
  T. Weber (moderator - UFRGS, Brazil)

Replicated data has been largely proposed as a feasible approach to reach dependability in distributed systems. But maintaining the consistency among replicated items is not as simple as it can be thought in a first moment. Data replication provides resiliency against failures but introduces new questions. What are the better strategies to manage the distributed replicas? How can a replica control protocol guarantee consistency without penalizing the system performance? How can the amount of dependability reached using replicas be measured? How can the cost of maintaining replicas be evaluated? Are replicated data feasible for new distributed system platforms as grid or pervasive computing? This panel aims to address those old questions discussing new strategies to create, distribute and manage replicas, the drawbacks of old and new approaches and also the benchmarking methodologies that could be useful to measure the dependability attributes of such systems.

Program committee

• Antony Rowstron (Microsoft Research Cambridge, UK)
• Bettina Kemme (School of Computer Science, McGill University, Canada)
• Emmanuel Cecchet (INRIA Rhône-Alpes, France)
• Fernando Pedone (École Polytechnique Fédérale de Lausanne, Switzerland)
• Jonathan Stanton (George Washington University, USA)
• Ricardo Jimenez-Peris (Universidad Politecnica de Madrid, Spain)
• Robbert van Renesse (Computer Science Department, Cornell University, USA)
• Svend Frolund (Hewlett-Packard Laboratories, USA)

Workshop organization

• Luis Rodrigues (Universidade de Lisboa, Portugal)
• José Pereira (Universidade do Minho, Portugal)

For further details, please send e-mail to wdddm@gsd.di.uminho.pt