Greg Secrist on ILM, and everything else under the "SUN"

The "Egg Basket" Idiom and Data Protection

My wife has several Longaberger baskets in a variety of shapes and sizes.  She uses some for floral arrangements, one for toting casseroles to the occasional dinner party, and others just for decoration.  She has baskets for different stuff.  If we still needed to carry our eggs from the hen house to the kitchen, I'm sure she would buy two more; one for the left hand and the other for the right hand.  Why? In case one slipped or tipped during the walk back, we'd at least have half of our eggs instead of none at all. 

Ok, you know where I'm going with this.  The proverbial "don't put all your eggs in one basket" is the perfect idiom to describe what not to do when trying to protect your data from loss or interruption.   It is a simple analogy, but I can't tell you how many times I see IT fail to apply it practically.

I've seen "tape only" backup systems being used as the sole container for data copies that are routinely recalled every time a user deletes an email or spreadsheet.   This means that every time an email mailbox is restored, a tape is mounted and read. Multiply this by dozens or hundreds of recalls a day and you can start to see why some recalls eventually fail.  (Especially when the system is architected with primarily capacity tape technology like LTO)  The result:  tape gets a bad rap because it is deemed "unreliable" and troublesome. 

I've seen customer's buy into the "hype" and trade out their tape-based backup systems for a VTL.  (An "all disks" egg basket)  Within a year, I see them scratching their heads why it is costing them 5-times as much to continue storing their data?  In one particular case, I saw an EMC EDL (EMC's VTL appliance) go down and fail to roll-over to the other VTL engine.  What happened?  This customer's entire backup system was down for several days, with some loss of data.  Why did it fail???? Over provisioning of virtual tapes and overload of the system.  The result: the customer is burned by VTL and it quickly becomes suspect for reliability.  Bet they wished they had a "last resort" copy of their data sitting out there on a tape somewhere!

This why you don't put all your data "eggs" in one data protection basket.  This is why I evangelize a tiered data protection architecture.  Call it tiers of defense.  If you only have a single wall around the castle, you are royally hosed if the enemy gets through a breach.  If you have an outer wall, a moat, and a final wall, you probably won't have to depend on that final wall. But it is nice to have just in case. 

Tape is your last line of defense that should be written too often and read from only when all other copies don't work.  Your first line of defense should be multiple point-in-time copies of your data on disk. The copy can be a database dump, volume split, snapshot, continuity volume, or any combination of these.  At the same time, you should make a point to make a "copy of the copy" to tape, just in case.  That makes tape your last line of defense that should prove reliable because you only go to it on occasion, not every time, every day. 

Think of it as splitting your eggs between baskets; the first being a Longaberger and the 2nd being a 9.99 special from Crate&Barrel.   

 

Posted at 10:46PM Jun 07, 2008 by Greg Secrist in Sun  |  Comments[1]