DAML DB

Home | About DAML | Announcements | Roadmap | Site Search

DAML DB is a prototype implementation of a scalable, high-performance persistent store for DAML content. It's intended to be embedded in other applications or tools.

Approach

The key ideas in DAML DB are:

convert all URIs and literals to integers, which are fixed-length and much more efficient to compare
store statements and other fixed-length information in memory-mapped files, which are persistent, can be indexed efficiently, and can grow to the maximum address space without consuming swap space
maintain linked-lists of statements involving the same subject, predicate, and object
use C/C++ rather than Java for added efficiency

Current Implementation

The current implementation uses

the Repat parser
Sleepycat Berkeley DB for bidirectional string-to-index mapping (I'm not in love with this implementation, but it's a starting point -- alternative suggestions would be welcomed)

Versions have been developed for Windows and Unix, and have been tested on Windows 2000, Redhat 7.0 Linux, and Solaris 8.

The current implementation uses 32-bit indices and pointers, which limits a database to about 25 million statements in a 1 GB mapped file (assuming some partitioning of address space by the OS won't make the full 4 GB available). A 64-bit implementation should avoid any practical limits for the foreseeable future.

API

The API is still evolving.

The primary interface is in the C++ header file IDamlDB.hh.

A Java interface is in DAMLDB.java.

Jena and RDQL interface

We provide org.daml.damldb.jena.Model, an implementation of the Jena com.hp.hpl.mesa.rdf.jena.model.Model and related interfaces, which provides read-only (currently) access to a DAML DB database. This can be used by itself, or with RDQL to execute queries against the database.

See the Examples section below.

Examples

We have DAML DB with RDQL executing queries on the collected results of the DAML Crawler, and plan to soon export that interface on this page.

Download

RELEASE NOTES

INSTALLATION (damldb, rdf2damldb, mysql2damldb)

Package	Type	Size	Date
damldb-0.1.0	.tar.gz	414,084	April 23, 2002
damldb-0.1.0	.zip	550,062	April 23, 2002
rdf2damldb-0.1.0	.tar.gz	253,819	April 23, 2002
rdf2damldb-0.1.0	.zip	295,133	April 23, 2002
mysql2damldb-0.1.0	.tar.gz		SOON

Possible Future Directions

provide an implementation for 64-bit architectures that effectively eliminates the current size limitations. Solaris might be an initial 64-bit target.
provide a server interface (possibly multi-threaded) to allow shared access from multiple clients
support concurrent access from multiple processes (probably single writer or multiple readers)
investigate alternative string-to-index mapping mechanisms
support RQL and/or other query languages
support some inference capability
...

Author

Mike Dean and Paul Neves

$Id: index.xml,v 1.9 2002/04/23 21:36:28 neves Exp $

DAML DB

Approach

Current Implementation

API

Jena and RDQL interface

Examples

Download

Package

Type

Size

Date

Possible Future Directions

Author