Poor man's server redundancy?

[oldtrlrnr]

I am helping a couple of friends who have small businesses, each with a single server. They have become concerned about the business cost of a server outage, but have a limited budget for redundancy. I used to work on this stuff, but have been out of it for several years, so I'd like to solicit suggestions from you all. Our goals are to provide quick & simple server failover at a modest cost, but not necessarily instantaneous or automatic. I've thought of a few categories of solutions:

1. Off-line server spare - Keep another server of same HW configuration; if live server, which has pair of HDD's in RAID-1 mirror, fails, remove one of the drives from the failed server, pop it into the spare server and fire it up, with same server name & IP address. I wonder if this would work - the disk controller in the spare server might not accept a drive from the other server. Also this would not protect from a major electrical problem, say both drives got zapped.

1a. A variation of server spare - maintain a spare server with distinct name & IP address, synchronize data nightly. If live server fails, shut it down, then rename the spare server and give it the IP address of the failed live server.

2. Failover SW such as Double Take or StorageCraft.

3. Server cluster - probably too expensive, especially if a SAN were used.

4. Virtualization.

Appreciate any suggestions.

Bob

P.S. This thread has been moved from Servers>General to ProLiant Servers (ML,DL,SL). -HP Forum Moderator

[waaronb]

The solutions to redundancy really depend on what kind of functions that server is performing.

For example, if it's a file/print server, then it's pretty easy. If it's a database server, then there are some options depending on what kind (Microsoft, Oracle, MySQL, etc). Web servers, same thing, there are options for that.

You also would take into consideration whether performance during an outage is expected to be exactly the same as during normal operation, or if some performance hit can be tolerated.

If a performance hit is okay during maintenance/failure, then you could setup a lower end system for backup purposes and keep the big expensive one for primary use.

And one key there is to find out just how much it would cost to have a server down for an hour, a day, a week, etc. because that will help dictate how much it's worth spending on the redundancy options. If the business cost of an outage is measured in thousands of dollars per hour, then it's a lot different compared to a few hundred bucks per hour of lost sales/productivity/whatever. Some small offices can probably get by just fine with calculators and typewriters if they really had to. I say that partly in jest, but partly serious. :)