| |||||||||||||
Future storage technologies and LinuxFuture storage technologies and LinuxPosted Apr 6, 2011 6:24 UTC (Wed) by butlerm (subscriber, #13312)Parent article: Future storage technologies and Linux It is not just random writes - without multiple track read heads, random reads will require three to five rotations, setting latency back perhaps thirty years. See here (pdf). Without a persistent non-volatile cache and extensive operating system support, it is hard to see how such drives could perform well even for general purpose desktop usage. For database and general purpose server applications, they would probably be out of the question unless the operating system could do far more than give the drive "hints" about what should be stored in low read latency and high read latency zones. However, for applications like DVRs and nearline storage, 5-10x the capacity seems like more than a worthwhile trade off for substantially increased access times.
(Log in to post comments)
5-10 times? Posted Apr 6, 2011 9:44 UTC (Wed) by tialaramex (subscriber, #21167) [Link]
Is this 5-10x the same as the 20% increase in the article? Or are they unrelated?
5-10 times? Posted Apr 6, 2011 15:32 UTC (Wed) by butlerm (subscriber, #13312) [Link]
The first 20% figure in the article refers to the gain from using 4K sectors. The second one I don't know about, but I assume it refers to the gain from using some form of first generation SMR, as opposed to the long term gains possible.
Of course I can't imagine why anyone would actually purchase (or market) an SMR drive with only a 20% capacity increase anytime soon given all the other problems.
TDMR vs SMR Posted Apr 6, 2011 23:52 UTC (Wed) by akanaber (subscriber, #23265) [Link]
The multiple passes (or multiple track heads) issue is with Two-Dimensional Magnetic Recording, which seems to be the next thing they want to do after "ordinary" SMR. The presentation goes on to admit "We really need that multi-element reader!" (page 52).
That was an interesting link, thanks.
TDMR vs SMR Posted Apr 8, 2011 22:27 UTC (Fri) by giraffedata (subscriber, #1954) [Link] The multiple passes (or multiple track heads) issue is with Two-Dimensional Magnetic Recording [TDMR], which seems to be the next thing they want to do after "ordinary" SMR. I think the point is that SMR isn't much use without TDMR because inter-track interference otherwise prevents you from realizing most of the density improvement that SMR is capable of giving. So it's right to lump them together and say the implications of TDMR are the implications of SMR.
|
|||||||||||||