String-Matching and Update through Algebraic Signatures in Scalable Distributed Data Structures

Abstract

Scalable distributed data structures (SDDSs) store large scalable files over a distributed RAM of nodes in a grid or a P2P network. The files scale transparently for the applications. The prototype system was designed by CERIA, experiments with this technology for Wintel multicomputers. The application may manipulate data much faster than on local disks. We present the functions we have put into the prototype we now call SDDS-2004. We improve the searches and updates of records in our SDDS files. An original property of these functions is the use of the algebraic signatures. This technique serves the distributed non-key record search. The search may concern the entire field or a (sub)string. The algebraic properties of the signatures act similarly to hash schemes in the work of R.M. Karp and M.O. Rabin (1987). In particular, sending a few-byte signature of the searched string alone, suffices for the search. This makes the communication between the SDDS client and server more efficient. It is also more confidential, since the signature in an intercepted message does not disclose the searched string. On the other hand, we use the signatures for the update management. The clients do not need to then to send updates which in fact do not change the stored records. Finally, our signatures help managing the concurrency control. We present our architecture and design choices. Performance measures validate our implementation. It is now available for download in site of CERIA