Performance of the SDDS LH*LH under SDDS-2000

An SDDS contains data records. Every record is identified by a primary key or by several keys. Records are stored in buckets at server nodes called (SDDS) servers. They are manipulated from client nodes called (SDDS) clients. The key value(s) determines the address of the bucket where the record should be. An SDDS file starts at one storage site. Inserts scale the file incrementally through splits. Each split transfers half of records at a server data towards a new server allocated to the file. Splits are not posted to the SDDS clients. Each client uses for the record addressing an image of the file structure. Client images may differ. In particular, splits may render an image inaccurate. A client may then make an addressing error and send a key search or a record to an incorrect server.