A Document Database Server: Yggdrasil
Yggdrasil is an electronic document storage and retrieval system that is under development. Primative elements it will store include plain text, editable document, scanned image, mail message, object code, digitized voice, and digitized video.
The basic data model is hypertext (derived from NoteCards). However, Yggdrasil integrates hypertext with naming and attributes (concepts common in file servers). Yggdrasil also supports containers (parent-child relationships), versions and alternatives, and indexing. The server design features online archival storage, client notification of significant events, large number of documents, very large documents, small documents, concurrency control, access list based protection, high performance, fast recovery, high availability, robustness, and data compression.
One of the goals of Yggdrasil is to make the server handle "large" amounts of data. Yggdrasil will have a storage hierarchy that will span main memory, rotating magnetic, and jukebox optical disks. The initial system will store about 100 Gigabytes, but the system will be designed so that it can support dramatically more storage and a variety of secondary storage devices.
The system has its high level design specified except for a high level query language. Implementation is proceeding by adapting part of the code from the rAlpine File Server. Yggdrasil will be written in Portable Cedar.
What Does It Mean for Xerox?
What are the Business Implications?
Xerox corporate strategy requires multimedia, large scale, unstructured, robust, and efficient servers for the storage and retrieval of documents. The Yggdrasil project addresses many of the research issues needed to provide such a service. This research is needed for Xerox to understand how to implement a filing strategy.
Yggdrasil is being designed so that different "protocol frontends" and be used. The frontends can present different interfaces to the server (e.g., XNS File Server). Yggdrasil should be configurable with various secondary storage devices to meet a variety of requirements for response time, storage capacity, crash recovery, availability, archival storage, backup, and robustness.
What are our Competitors Doing?
Much of the research in database systems is involved with increasing the semantics of the system. Examples of this work is extensible and object-oriented database systems.
Another competors are Image filing systems. They are highly structured. Information is ususally indexed by one key or indexed via a relational database system.
A few other hypertext based servers are in development. The major ones are at MCC and at various universities.