Disk scheduling policies with lookahead
Disk scheduling policies with lookahead, A. Thomasian, C. Liu, ACM SIGMETRICS Vol. 30, No. 2, September 2002, pp. 31-40.
Disk scheduling methods that we might already know are concerned with minimizing the seek time, for example, the FCFS and the SSTF methods. However, the summation of both seeks and latency time is more preferable in modern disk. Therefore, the authors introduce some new disk scheduling methods. For example, the SATF policy which takes into account the sum of seek time and latency time is therefore preferable.The authors review the major disk scheduling methods such as FCFS, SSTF, CSCAN, CSCAN-Lai, SATF, SATF, HOL and SATF-RP. They describe the simulation model used to evaluate the relative performance of the disk scheduling methods, and analyze the simulation regarding to those methods. The main contribute is that they extended CSCAN and SATF with look ahead to be able to cope with the dynamic nature of arrivals to the system.
As we might know, we don’t concern a capacity of disk as a major issue like before, and the speed of the seek time became much faster than before. I believe a disk scheduling method is suited for some specific data, it seems to me like there will not be such a method that can optimize all data which is stored in the disk. My question is that they should have a disk scheduling method which acts like the MTLQ (Multi level queue) that we have studied in the early chapter, where we could select right algorithm and move up and down depends on the starvation level. That should be very more interesting.
In my opinion, the read and write speed could improve by increasing speed of motor and some more mechanical stuff rather than using scheduling methods, of course there would be some improvement but only minor, since today we don’t feel that the bottleneck of transferring data is occurs at memory device.
For this paper, I had like to rate the significance of this paper as 3/5(modest), because 20% of the paper review the scheduling methods which most of us already know, the simulation doesn’t show us a significant result of improvement of disk utilization, and this should be the most noticeable deficiency of the paper.