Re: Database writing architecture

In article <316E92EE.41C67EA6_at_exu.ericsson.se> Gerald Ward <exuward_at_exu.ericsson.se> writes:
>JB,
>
>That would explain the raw partition writes, and I agree
>with you that on raw partitions both oracle and sybase
>take over the writing to disk. If you've looked at linux
>code, there are actual scheduling algorithms that depend
>on the location of the disk's actuator arm.
>
>On a filesystem however, their higher level arcitectures
>come into play (at least I think so). My theory is that
>Sybase synchronizes the writing of the db-blocks and the
>writing of the log-blocks and that Oracle does not.
>
>If you only write the logs and can reconstruct the db-blocks
>from only the log blocks, then you have the ability in
>a power failure to reconstruct the db-blocks. In normal
>operation, the db-blocks should be written a significant
>amount of time AFTER the log blocks to avoid data corruption.
>
>If sybase writes the db blocks and log blocks at near
>simultaneity, then corruption can happen with power
>failure. My question is... Does this indeed happen?

Since I've seen corruption without power failure, I would think Sybase could have corruption with power failure. I've also seen it just plain ignore commands to switch db's. Makes for real interesting DBA script automation.

>
>
>--Jerry.

-- 
Joel Garry               joelga_at_rossinc.com               Compuserve 70661,1534
These are my opinions, not necessarily those of Ross Systems, Inc.   <> <>
%DCL-W-SOFTONEDGEDONTPUSH, Software On Edge - Don't Push.            \ V /
panic: ifree: freeing free inodes...                                   O