Hi Marc,
We've had the same questions and cam up with the following plan:
We'll setup a server with raid 5 (linux). With plenty disk space for this
year to come (estimated).
Then the next year we'll buy a new server..... with raid 5 and enough disk
space for the first year.
Patches new releases are taken into account in this new installed server.
The backup procedure will be tested getting the data on the database.
Then a test period follows before the lot goes live.
I know I won't win a price with this idea but specially the test possibility
allows me to sleep (at night) when changes on the database have to be done.
And it might seem expensive but considering all options this might turn out
to be the cheapest solution after all.
Cheers Wim.
>Yet again, I'd like to bring up this subject from last December. There
>seemed to be an important point which was overlooked: scalability.
>
>Being not so knowledgeable on the subject of raid or any array of disks, I
>went to the freebsd-scsi and -fs mailing lists to find out about expanding
>disk space. Unfortunately, all I could find was that solaris could do it.
>
>Now I'm back to square one, looking for a storage solution for my
>postgresql db. More specifically, something scallable and very fast. In my
>reading, I found that raid5 could also fail and that a tape backup is
>always recommended. Therefore, I would think raid 0 or 1 or 0+1 could do
>just fine having a tape backup as fault tolerance. As for a software or
>hardware implementation, opinions vary. Marc G. Fournier seems to prefer
>the software solution while there are many people preaching the benefits
>of hardware raid out there. But the question remains, how can all this be
>scallable, ie how can I expand on existing storage space.
>
>Looking forward to your replies,
>Marc Tardif
>
>
>************
>
