Abstract: Scalable Distributed Data Structures (SDDS) is a file of records distributed to servers of a multicomputer. Faults of record placement in SDDS file may lead the file to crash. In this paper a fault model for decentralized SDDS LH* (DSDDS LH*) is given and, for such a model, a fault tolerant record placement architecture is presented. Client message forwarding is supplemented with client message backwarding. Split token (used in DSDDS LH* for controlling the file scalability) is supplemented with token tracing protocol and so called token level. These additional mechanisms allow for regeneration of DSDDS LH* file damaged by transient control faults.
B I B L I O G R A F I A1. Dongarra, J., Sterling, T., Simon, H., Strohmaier, E.: High-Performance Computing: Clusters, Constellations, MPPs, and Future Directions. IEEE Computing in Science and Engineering (March 2005)
2. Litwin, W., Neimat, M.-A., Schneider, D.: LH*: A Scalable Distributed Data Structure. ACM Transactions on Database Systems ACM-TODS (December 1996)
3. Litwin, W., Neimat, M.-A.: High-Availability LH* Schemes with Mirroring. In: Intl. Conf. on Coope. Inf. Syst. COOPIS 1996, Brussels (1996)
4. Sapiecha, K., Łukawski, G.: Fault-tolerant Control for Scalable Distributed Data Structures. Annales Universitatis Mariae Curie-Skodowska, Informatica (2005)
5. Sapiecha, K., Łukawski, G.: Fault-tolerant Protocols for Scalable Distributed Data Structures. In: Wyrzykowski, R., Dongarra, J., Meyer, N., Waśniewski, J. (eds.) PPAM 2005. LNCS, vol. 3911, Springer, Heidelberg (2006)
6. Łukawski, G., Sapiecha, K.: Software Functional Fault Injector for SDDS. GI-Edition. In: ARCS 2006. Lecture Notes in Informatics (LNI) (2006) 
Pełny tekst 
DOI 
Web of Science