jgh@root44.UUCP (08/14/87)
[Foolish ideas dept] In article <2838@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: > So, why not put the disk in a vacuum? The problem is, of course, flying the heads. I was batting this idea around with a lunatic friend last night in the pub, and we came up with the following: Build the disk platters as a sandwich of (from the outside) oxide (plated, sputtered, whatever), metal (strength member), superconductor. Since liquid-nitrogen class superconductors are due Real Soon Now, this won't be too expensive :-) Replace the flying-shape part of the head with a (relatively) large, low strength magnet. Strong enough to float over the superconductor, weak enough not to affect the data. Low cost versions use permanent magnets and never land on the platter even without head lifters. High performance version use electromagnets for low head mass. While you're at it, use maglev bearings for the spindle and head-arm. Opinions, anybody? Jeremy -- Jeremy Harris jgh@root.co.uk
johnw@astroatc.UUCP (John F. Wardale) (08/17/87)
A few comments: It is best to run disks in posivite proessure to keep out unwanted "crud". Flying the heads in H2 or He would get you (maybe) twice the speed (of air) before you hit mach. Reducing pressure gets you less than this. Some (fuji I'm fairly sure) makes Parrallel Transfering Disks that get 10 to 20 mega-bytes per second. They are head/surface, like normal disks, but there's a copy of the tranfer logic for each one! Benchmarks I've seen/heard about (multi-user unix usage patterns) indicate that you lose about half of this, but at 5 or 10 times rates, you gain 2.5 to 5 times the "effective" transfer. [ Speculation on the lose: the heads are fixed together, and you may not always want the "disk-cylendar" so the seek-delays hit you twice as hard. ] John W - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Name: John F. Wardale UUCP: ... {seismo | harvard | ihnp4} !uwvax!astroatc!johnw arpa: astroatc!johnw@rsch.wisc.edu snail: 5800 Cottage Gr. Rd. ;;; Madison WI 53716 audio: 608-221-9001 eXt 110 To err is human, to really foul up world news requires the net!
hank@spook.UUCP (Hank Cohen) (08/19/87)
A further note on the Fuji parallel disks is that they require you to do full cylinder transactions. The disks are also formatted for very large sectors like maybe 1 or 2 per track so the UNIX file system dosen't map onto them very nicely.