On 02/06/2007, at 11:30 AM, Mario Ruiz wrote:
Hi Simon,
As Peter suggested, we also believe that the backup function is
more that just storing the stuff somewhere out of the practice, or
even out of a device. Its the complete backup && security &&
recovery lifecyle of a practice dataset.
In our case, Insight provides a single click GMAIL Backup for a
practice. GMAIL also provides around 2Gb of free storage, which in
the case of Insight holds several hundreds days worth of backups
for a dataset spanning a couple of years of a practice trading.
Insight backups are strongly encrypted to ensure privacy of
practice data and avoid content parsing by bots. When you run out
of GMAIL space, you do a bit of cleanup and delete the oldest few
weeks (it depends on your strategy) and off you go again. Simple
and reliable with very short upload/download times.
One advantage of using GMAIL is that you can get to your backup
from anywhere and download your latest backup. Then you can use
"Insight Anywhere" (SSL) option to restore your system remotely. No
need to restart you remote machine.
Regarding your "Tip", I believe that risking your business
continuity for the sake of a few dollars a month might be unwise,
the upload cost usually is no more (perhaps Telstra is a lot more)
than an additional three or four Item 23 a year. Surely a practice
can afford that, otherwise go paper.
For several of our sites, we run an rsync over ssh backup, stored on
our servers, but this is never a principle backup.
For a couple of others, the same thing happens to their home machines
and is restored automatically (except in the case of MD, where it's
almost automatic).
When we are storing client data, it is 2048bit GPG encrypted with the
clients key. We also use our key with the practices consent, in case
theirs is lost.
This works really well. Once setup properly it doesn't fail. We do
test restores for the sites in-house. Provided the data is encrypted
prior to transport, it could go anywhere that you can get it back
quickly. In the event of a failure at one of our local sites, it
would be onto a removable hard disk and into a car minutes after the
failure.
This obviously doesn't scale so I have no ambition to turn it into a
product. I can think of many easier ways to make $5 a month.
Telstra charge full price for uploads once you are over quota, and
last time I looked, offered no throttling plans. So this would be
inconceivably expensive if you are uploading a full database snapshot
each night. Unfortunately due to use and indexing, rsync has little
effect on reducing the size of these files. A database that can
backup incrementally would be much more efficient.
Also one needs to count the cost of recovery of more than just your
database. A server failure usually means an install onto a PC to keep
the site working, having to reconfigure all clients, then a repair of
the server and doing it all over again. We seem to run into an
average of one of these per year and the time taken has been over 20
hours in tech time on each occurrence (each has been large, Active
Directory domains). To avoid this kind of recovery we're starting to
insist on nightly snapshot backups of the entire server to a
removable hard disk as part of the backup strategy, and allow a
machine on-site that has similar if not identical hardware to the
server. This will ensure that in 90% of failures we can have the site
back up and running very quickly.
cheers,
Peter.
_______________________________________________
Gpcg_talk mailing list
[email protected]
http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk
