rdt@houxk.UUCP (06/08/84)
I am pretty naive about disk controllers and looking for general information. I understand that all disk drives tell their disk controllers when they reach the first sector within a given track. Do any drives tell their controller when they reach the start of any sector within a given track as well? This could be useful to speed up disk access if a disk cache scheme is used. As soon as one gets to a given track, the controller could allocate buffers of size sector length. As one passes the position of the next sector on the given track one could fill buffer 1, and then sector buffer 2 and so on. As the signal for the first sector within a track is detected. One could remap the sector buffer index names to track sector numbers. Assuming that a simple subtraction is necessary to point to the desired sector number within hte track, the worst case sector rotational delay is reduced from 2 rotations to 1 rotation plus some fast diddle time that is neglible. Is this done in any of the current controller/drive boards? If not, why not. Is it hard to encode the start of any sector in any given track with 1 "magnetic stripe" type and the start of the first sector in any given track with another "magnetic stripe" type? If it is hard, is it a problem in that no interface signal exists or is it a disk media issue where there isn't enough context to determine the start of any sector within the given track. Pretend the picture below is a disk. Graphically, I mean the following: 2 3 A B \ / / \\ <- this is the first sector stripe; the rest are D C for the begining of any sector in the track. 1 0 Sector Buffer Index name: A B C D (order of sector start seen) Sector Buffer Address 0 1 2 3 (established after first in track seen) Association of Index to buffer name is very fast wrt rotational latency. Lastly, is the average time to position a single head 5.25" winchester disk at the right radius (step time, seek time) long wrt the time for the average time to rotate the average amount once in a track (I guess 1/2 the rotational latency)? I thought so since 20 ms access is average and 1/3600 rev/sec is .03 ms. Even if I'm wrong about the rev/sec and it is 1/360 -> 0.3 ms; this seems a short contribution to hte overall access time. Is there literature on this subject? Richard Trauben AT&T Bell Labs