A modest Linux USB suggestion
Submitted by srlinuxx on Mon, 08/25/2008 - 17:05.
Did you know that there are two basic kinds of USB 2.0 drives? I didn’t. But, now thanks to Robert L. Scheier’s article, Not all USB drives are created equal, I now know that are significant differences between drives. And, in particular those differences matter a lot to live USB capable Linux distributions like Fedora 9.
The differences, in short, is differences in the memory type and their I/O controllers. The results are anything but trivial. One type of USB drive will run two to three times faster than their slower brothers and, potentially, last 10 times longer.
As Scheier explains, “The single biggest factor in USB drive performance is whether it contains one of two types of memory: SLC (single-level cell) or MLC (multilevel cell). SLC stores one bit in each memory cell, and MLC stores two bits in each cell.” The more expensive SLC memory runs twice as fast as MLC, “with maximum read speeds of about 14MB/sec. and write speeds of about 10- to 12MB/sec.” It also lasts much longer.
MLC and SLC are the same ? MLC needs multi-DSPs ?
SLC existed for many years already. The mean time between failure increase from 8000 hours to 230,000 hours is due to a surface preparation on magnetic media.
The same life expectancy is expected with MLC with the same surface preparation to prevent head crash mechanically on the magnetic media(scratches).
The difference is multi level magnetic tracks use different bias frequency(below 70khz and above 90khz) to penetrate deeper into the magnetic media and has to read out with that frequency. So, MLC needs two DSPs to run at the same speed as SLC. But realtime 2 data channel mixed signal drivers will have twice the data speed than SLC without changing RPM spindle speed.
We proposed MLC to Seagate, years ago. Now we are proposing two DSPs to speed up data transfer to two out of 127 USB ports. Someday, multiheads will bewithout magnetic free spacing between tracks; when multi level bias frequencies fill up the entire magnetic fild. One track with multi bias frequencies is different from the very next track, using two heads on each arm. Have fun. data density will double or triple, again and again. Giant magnetic resonance readout and coil recording heads will have to be more and more narrow.