On Fri, 16 Jun 2000 18:43:42 -0700, Jason Wharton wrote:

>> First the problem, then the solution.
>
>I thought I had made the problem sufficiently clear. I'll recap and extend
>it some to clarify.
>
>Requirements:
>Huge database. (many GB worth of data)
>Database needs to stay on-line 24X7.
>98% static & generally inserts only.
>In case of failure, downtime needs to be minimized.
>Immediate failover isn't critical.
>Cannot lose over one day's work but up to that is tolerable.
>All input sources are available for two days to be re-entered if necessary.
>
>Problem.
>Obtaining an efficient backup on a regular basis.
>Efficient meaning the total I/O and CPU toasted in the process, among other
>things.
>
>Problem extension #1:
>Same huge database.
>Needing to be sync'd with many remote copies on a daily basis.
>Real-time is not necessary.
>Remote copies are for inquiry only.
>Potentially low and/or expensive bandwidth constraints.
>New base versions are express-mailed via DVD media or DAT.
>
>Problem extension #2:
>Same huge database.
>DBA needs to cancel out an entire days work due to certain problems
>encountered.
>Database needs to stay on-line.
>
>Have at it. That should cover most of it.
>

Before I start to say BS, can you all answer a question to
me to help me to build my theory and express my idea?

The question is: is IB transactions numbers (all of them, does
not matter what is done within) unique between 2 full backups?
Unique in dbwide way.

Transaction numbers can be reassigned by normal garbage
collection? Let's say : T120 will be reassigned after GC does it
work and don't rest no rec with T120 ( all of them were deleted)?

[]s Fabricio




[]s Fabricio
Delphi C/S Developer