SOHO backups...

I have already posted this in some other forums/newsgroups but knowing some of the population here, I figured I might get some innovative ideas...

The question is "How to backup a SOHO (Small Office/Home Office) environment???"

The basic parameters are a situation where the company in question has 3-5 TB (3,000-5,000 GB) of data on one or more servers.....

Larger companies can use off-site (co-lo) servers to maintain secure backups, with multiple staggered locations...

"Personal" systems rarely get backed up in a manner that is approriate for a business.

For example, restore the state to end-of-day for any day within the past 3 months, restore the state to end-of-week for a year, restore the state to end-of month for 2-3 years, restore the state to end-of-year for 5 or more years....

Disk drives are currently "cheap" with costs averaging about $0.10/GB, but have their own issues.....

HIGH capacity TAPE systems have a large cost for the "infrastructure"...

So how do people here which have these types of systems (or support customers with these types of systems), accomplish a robust system backup plan?????

The system we use at the office is far from high-tech. We have an image disk with the operating system part of each server, so we can easily setup a new one. The data on each server is backed up to a backup server during each night. Important data must therefore always be stored on one of the servers and not on the HD of one of the office computers. This ensures that all our data is duplicated on the backup server.

Each friday the backup server copies all data to a large removable harddisk - in encrypted form(!). This harddisk is taken home by the owner of the company who on Monday will bring a different harddisk back. He has copies of the image disks for the server at his home too of course. We replace each of these drives at least once a year.

I regard keeping backups off-premise as important as the actual backing up. If the company burns down then we will still have all our data so we can immediately setup shop somewhere else and continue doing business. Keeping your backups at the same location as the servers does not help you in case of such a calamity. It did not take me long to convince the owner of this. Every once and a while he gives me a disk home as well just to ensure there are copies at different locations.

Each server has a redundant RAID system to protect us from single drive failures. So far the backup server only needed to rescue us from human error - in error deleted folders and such. The only other times we used it was for test. When we replaced two of our servers a couple of months ago we decided to do a restore of our backup to the new server, rather than the usual copying of the data. This worked flawless. Again this is very important as a backup you can't restore is like bicycle for a fish - useless.

This system as such do not give you any specific options for going back to a particular date, but we don't really need that.. The main part is that it is mostly automated and therefore cannot be forgotten.

Edit: One reason for going for hard disks is that you do not need any specific hardware to start restoring computers. With tape systems you always need to ensure that you have a spare tape drive off site, which needs to be tested regularly so that you know it will work in case you need it. Otherwise you may find a compatible tape drive hard to find after a major catastrophe hit the company three years down the line. A backup medium without a compatible reader is also pretty useless...

Edders,

Thank you for posting some very good information; especially about the need for off-site.

As you indicated "user errors" (deletion of files, etc) are different than "crashes". One of the special concerns is that the problem may not me noticed for quite some time.

Consider a user who has to produce an end of year "report" (meaning some type of document, not a query against a db). The information will be basd on the contents of the 3 quarterly reports. which are now 3,6,9 months old respctively.

With a "short rotation" policy, it is discinctly possibly that the first two were destroyed in July, and by December/January all of the backups have been cycled.

To counter this, one can set up a hierarchical rotation (every day for some # of weeks, every week for some # of months, every quarter for some # of years), but this increases the media count significantly (which increases costs and physical storage requirements).

It is for these later cases the I find the chalenge in arriving at a good solution for multi-terabyte servers. Disk HAS dropped tremendously in price, but more than once I have had a "Book" type drive be damaged in transport (and the damage not known until an attempted restore/reuse).