Random Access Tape 2448

Michael Wojcik

I'm going to let you worry about the implications of agreeing with an butt; (mostly) I'm just arsing around because I stubbed my toe on your pedestal.

But lets' do some math...

--------------------------------------------------------------------- *Comparison of one HD Cylinder to a Bi-directional Positioning Tape* (spooling-cacheing doesn't count for obvious reasons, and amount of data is the same for reasons of comparison)

Random Access:

cylinder: average 1-2 the media needs to be traversed to get to a random point (1-2 a rotation in one direction)

tape: average 1-2 the media needs to be traversed to get a random point (1-2 tape movement in either direction)

Sequential Access: cylinder: average 1-2 a rotation or half the media. tape: immediate.

So, in this comparison they're the same for random access and tape has an obvious advantage in sequential access.

---------------------------------------------------------------------- *Comparison of a MultiCylinder HD to a Bi-directional Positioning Tape* (presuming a 7200rpm drive with an avg track-to-track of 10ms)

Random Access

HD: average track-to-track is 5-6 of a rotation + a further 1-2 rotation to position on the track, so the entire media has to be cycled an average of 1.33x to get to a random point.

TD: see above... 0.5x on average

(and an interesting sidenote: even if you gave each track it's own head

----------------------------------------------------------------------

(thanks: for some reason I couldn't think of the word "topology")

I tell a machine "give me record X" and it gives me "record X", it's random access; I don't care if it's a card-reader with an autocache. Yes this is a programmers POV... but the other POV's have no bearing in the "real world". Likewise the speed issue is only relevant on an individual case basis.

rpl