|
|
|
|
|
|
|
12. A block read request is sent to an Hitachi DK516 disk for 16 sectors, beginning with surface 3, track 23, sector 7. At the time the request is received, the active head is positioned over the start of surface 12, track 1, sector 0. If the transmission rate to the buffer is 3 MB/sec, what is the elapsed time until the data are in the buffer? |
|
|
|
|
|
|
|
|
13. For study 9.2(a) and (b), what is the perceived delay per request? |
|
|
|
|
|
|
|
|
14. Repeat study 9.2(a) with an 8-kB block. Assume that with an 8-kB block the processor makes a request every 700K user-state instructions. Which block size provides better user performance? |
|
|
|
|
|
|
|
|
15. Repeat study 9.5 with 50% 1-block accesses and 50% 16-block accesses. |
|
|
|
|
|
|
|
|
16. A RAID storage system is to be implemented with 16 drives. These drives can be configured independently, striped, or synchronized. Configure a RAID-1, RAID-2, and RAID-5 storage system that maximizes the amount of data storage. What fraction of storage is devoted to data for each configuration? |
|
|
|
|
|
|
|
|
17. The processor in study 9.2(b) is replaced with a 100-MIPS processor. Assume the memory is increased to support multiprogramming of degree 6 and the disk accesses are uniformly spread over four disk servers. Compute the achieved throughput and individual server occupancies. |
|
|
|
|
|
|
|
|
18. A disk cache buffer is added to the processor in study 9.2(a). Assume a miss rate of 30%. Estimate the effect of the buffer on throughput by modifying the effective user and system time per physical disk access (miss). Assume Tsys = 0 on a cache buffer hit. |
|
|
|
|
|
|
|
|
19. An additional 1 MByte memory is to be added to the processor in study 9.2(a). It can be configured as a disk buffer cache (25% miss rate), or as main memory to increase the degree of multiprogramming to three, or the block size can be doubled (thereby doubling the user instructions between disk references). Which option provides the best user performance? |
|
|
|
|
|
|
|
|
20. Repeat study 9.3 with Tdisk=35 ms for n=1 and n=10. |
|
|
|
|
|
|
|
|
21. For the configuration in study 9.3, assume we distribute the disk storage throughout the workstation network. Each workstation gets a single disk with Tdisk= 40 ms. Disk accesses are distributed uniformly over all disks in the system. Assume that each access requires 20 ms of system processing on the local processor and 20 ms on the target processor. Assume access to the local disk incurs the same network cost as a remote access. Estimate the achieved throughput and the processor and disk utilizations. |
|
|
|
|
|
|
|
|
22. Repeat study 9.3 with the following modifications: Assume each work-station has a local disk. Assume 50% of user requests are to the local disks. Model the effects of the local disk by modifying the effective user and pause times per server access. |
|
|
|
|
|