This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: CS-350: Fundamentals of Computing Systems Page 1 of 11Lecture Notes © Azer Bestavros. All rights reserved. Reproduction or copying (electronic or otherwise) is expressly forbidden except for students enrolled in CS-350. Stateful Resource Scheduling So far, we have only examined scheduling approaches for state-less resources. Recall that we defined such resources as those whose service time (for a given request) is independent of history. Recall also that we have noted that this assumption may not be correct for many resources, in that the ordering of requests could impact performance. In this chapter, we look more closely at one such resource—namely the disk. We look at possible disk scheduling techniques, noting that, as with CPU scheduling, our coverage of disk scheduling is intended to exemplify the choices and basic “tricks” we may employ to deal with other stateful resources. Examples of other stateful resources include memory systems (due to caching effects), sensors such as cameras (due to calibration overhead), and actuators such as elevators, conveyor belts, etc. (due to spatial dependencies). Disk Service Time Before we delve into specific disk scheduling techniques, let us first consider what contributes to the service time for a request submitted to a disk. Broadly speaking there are three main components: (1)Seek delay: This is the time it takes the disk head to move so as it is lined up with the cylinder of the disk containing the block of data to be read/written. Seek times vary quite a bit (depending on technology and size of disk). In the worst case, the seek time would be proportional to the diameter of the disk. Broadly speaking, it is typically in the range of 2-10 milliseconds. Saying that seek time is proportional to the distance traveled by the head is really an approximation. In reality, seek time is highly non-linear, due to acceleration, as in the second law of Newtonian physics, as well as other characteristics (such as mechanical properties of the disk). For the purposes of this course, we will assume that seek time is simply proportional to the distance traveled. (2)Rotational delay:This is the time it takes the disk to rotate so that the sector (on the cylinder) corresponding to the block of data to be read/written is under the disk head. In the worst case, this is the time it takes the disk to make a full revolution. Again, rotational delays vary quite a bit (depending on the speed of disk rotation, which is related to size and power consumption, among other considerations). Broadly speaking it is typically in the sub-milliseconds range and is linear, since the rotational speed is typically constant. (3)Transfer time:This is the time it takes the block to be read. Typically this is a constant (related to the angular velocity of the disk rotation) and is quite small—clearly much smaller than rotational delay as a cylinder may have a dozen or two sectors....
View Full Document
This document was uploaded on 04/14/2011.
- Spring '09