On Jan 17, 2011, at 10:45 PM, Jack Coats wrote: > In a quick spreadsheet, one years backup of 1G, that expands at > 15%/yr, and has a 5% daily change rate has the one gig, grow to almost > 21G in just one year. This assumes 100% retention, and linear growth > during the year. I can see this is an argument for tape instead of > only to disk!
A great deal depends on the timescale you look at and how much that data is compounded. As you look at larger timescales (days instead of hours or minutes), a lot of those smaller changes fall away, because you don't need to store the changes from A -> B -> C -> D -> E -> F -> G -> H -> I -> J, and you can instead just store the changes from A -> J. As you can roll up more of those changes and you don't need to store all the atomic data points in between, you get what you could call temporal data compression, or perhaps more accurately temporal data de-duplication or maybe temporal data elimination. So, a daily backup scheme would have one set of percentages that is compounded daily, while an hourly backup scheme would have a presumably lower set of percentages that is compounded hourly. And the hourly compound results could well be much, much larger than the daily compound results, far beyond what the difference in time scales would otherwise seem to indicate. Of course, CDP would have it's own set of formulas and apparent growth patterns, even if the actual dataset is exactly the same in each case. Unfortunately, tools like find(1) make it relatively easy to see what the list of files are that have changed on a daily basis, but they don't necessarily make it so easy to see the list of files that have changed on an hourly basis. And all you know is that there have been changes, as opposed to any indication of what the changes are. -- Brad Knowles <[email protected]> LinkedIn Profile: <http://tinyurl.com/y8kpxu> _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
