[LINK] Web Central outage - was The Sky didn't fall. Yet.

[Scott Howard]

On Wed, May 6, 2009 at 4:32 AM, Brenda Aynsley <bpa at iss.net.au> wrote:

> I think I would put a lot of stock into that sentence (yes it is a
> single sentence) alone and if my ISP was so large and the volume so
> great that this was true AND they had not implemented a way of managing
> this appropriately, I wouldn't want to be continuing with them.


Backups for mail as a disaster recovery mechanism is always a difficult
concept, and one you never really want to have to use.

Mail data, especially for an ISP who is providing POP mail to most of their
customers (as opposed to server-based protocols like IMAP and/or Exchange)
is very transient.  Some data will remain on the server for only seconds,
and depending on the ISP and the customer base the half-life of all mail on
the server is probably less than a day, and definitely less than a few days.

Backups will obviously take a point-in-time copy of the data, which means
that if you do need to restore not only will any mail that came in after the
backup be lost forever, but also anything which was already downloaded by
the client (or deleted, if they are using IMAP/etc) will suddenly re-appear
for a second time.  Actually fixing the mail store issue, if possible, is a
much better soltution and I'm not surprised they went down this track.

Many (most?)  ISPs used to simply take the approach that it was too
transient to even both trying to backup, and thus didn't.  In a world of
webmail and IMAP (both of which can result in the mail living on the server
for longer) that may have changed a little, but I suspect there's many ISPs
that still don't even bother with backups of users mailboxes.

Add to that the fact that backups systems - especially fast ones - aren't
cheap, and I'm not surprised that it would have taken them a long time to
restore an entire mail store.

 Scott.