On Wed, Feb 19, 2014 at 06:10:25PM -0500, John Stoffel wrote: > > Ok, I bit. What does "Mean time to Data Loss" really mean? It's just > a graph showing that two lines go down and the number of drives goes > up. They both go down. But this useless marketing article never says > why. > > So I went and found the paper, which is here: > > https://www.usenix.org/legacy/events/hotstorage10/tech/full_papers/Appuswamy.pdf > > for your reading pleasure. I also note that the same hot topics > conference has the paper: > > Mean Time to Meaningless: MTTDL, Markov Models, and Storage System > Reliability > Kevin M. Greenan, ParaScale, Inc.; James S. Plank, University of > Tennessee; Jay J. Wylie, HP Labs" > > which you can find here: > > https://www.usenix.org/legacy/events/hotstorage10/tech/full_papers/Greenan.pdf > > This article makes a good case to me that MTTDL is useless. Not that > I think their normalized time to dataloss metric is any more useful.
That article [Greenan] objects to the use of Markov models for computing MTTDL, not the metric itself. As the article points out, Monte Carlo simulation is also used to compute MTTDL figures in relative time, which obviates the shortcoming with Markov methods for this purpose. -- Charles Polisher _______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
