Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Anticipatory Disk Arm Movement

IP.com Disclosure Number: IPCOM000036173D
Original Publication Date: 1989-Sep-01
Included in the Prior Art Database: 2005-Jan-28
Document File: 2 page(s) / 70K

Publishing Venue

IBM

Related People

King, RP: AUTHOR

Abstract

Algorithms are disclosed for the reduction of the average distance which disk access arms must move in response to a series of requests. This is accomplished by the movement of the access arm during periods of idleness in anticipation of future requests.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 60% of the total text.

Page 1 of 2

Anticipatory Disk Arm Movement

Algorithms are disclosed for the reduction of the average distance which disk access arms must move in response to a series of requests. This is accomplished by the movement of the access arm during periods of idleness in anticipation of future requests.

Suppose requests to the disk are independent of each other, and uniformly distributed over the N cylinders which the disk arm accesses. Normal FIFO (first- in, first-out) scheduling of the access arm yields an average seek distance of N/3. Whenever there are no requests waiting for, or receiving, service, move the arm to cylinder N/2. The next request requires an arm movement which, on the average, is only N/4. This is a 25% reduction. For an access arms which is busy 10% of the time, this savings is seen by 90% of the requests.

Suppose that, rather than complete uniformity, there is a hot spot at cylinder L whose request probability is p. (Without lack of generality, assume L < N/2.) If the arm is placed at x, then the expected seek distance to the next request is E =p(x-L)+(1-p)(N/2-x+x*x/N). The best choice for x is the one which minimizes E, which can be found by taking the derivative of E with respect to x, setting that equal to zero, and solving for x, which gives x=N(1-2p)/(2-2p). In general, the distribution of requests is used to find the expected seek distance given the choice of x, and then x is chosen to minimize that expectation.

Another generalization is to consider di...