The VDB Toolkit is a new database toolkit under development at Daylight.
It provides client-side, readonly access to multiple Thor
This toolkit supports the vdb
object, which behaves as if it were a single database containing all the
data in it's member databases.
The raison d'être of vdb's is to simplify access and management of
large database projects, such as VCS
which is expected to grow to ~70 databases).
In such cases, it is much more convenient to manage individual databases
as logically independent entities yet be able to access all the data as
if it were a single database.
Such situations probably occur in many large institutions, e.g. managing
data in corporate, project, and commercial (but in-house) databases.
We're considering offering the VDB toolkit as a product for general use.
- vdb objects act like regular Thor databases.
One new object type and one new (non-polymorphic) function are added
- Logically, vdb objects are sets of real database
objects, and are managed with the normal object set operations
dt_member(), dt_add() and dt_remove().
- vdb database objects are readonly.
- The vdb toolkit handles tasks such as resolving datatypes
and merging datatrees.
- Client programs will be upgraded to access vdb's
(e.g., xvthor and xvmerlin).
- vdb's provide a method of accessing really large datasets.
- If a vdb has a lot of member databases, and each has it's own read
password, security can be a nightmare.
- dt_stream(vdb, TYP_DATATREE) is difficult to implement
- We need to provide a mechanism for merlin access of vdb's.
(Solution will probably be to make merlinserver be a vdb client rather
than implementing an internal "virtual pool" object.)
- A method for parallelizing retrieval has yet to be figured out.
This is more an issue for use on WANs than LANs.
vdb's are definitely in our future.
However, it is not clear whether the vdb toolkit will make the
cut for Release 4.71.
Daylight Chemical Information Systems, Inc.