Lester Caine wrote:

>I still prefer files to blobs, for exactly the above reason. The bulk
>static data can be stored in several sites, if the database throws a
>wobbly, there is only a small area to fix, and one can get running
>quicker than if the whole thing has to be reloaded. Once the core system
>is restored, the fine detail can be loaded as required. If the system
>structure is built correctly, the end users need never know that a
>machine has failed?
>
>
>

I have probably mentioned it earlier, but Netfrastructure has a feature
called a blob repository. A repository is a separate set of database
files for holding blobs. A field may name a repository, and any number
of blob/clob fields in any table can share a repository. The repository
definition contains a pattern for determining repository path names,
which by default use the database path itself, the schema name, the
repository name, and file sequence number. Repository file rollover by
size, time period, or a combination of both. Care is taken in the
implementation so database replicants have functionally identical, ie
interchangeabout, repository files.

The primary benefit of repositories is that large stuff is stored in
generally stable files, drastically reducing the size of the primary
database. This has a huge effect on backup, restoration, and failover
times.

--

Jim Starkey
Netfrastructure, Inc.
978 526-1376