Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Storage (cont’d) Disk scheduling Reducing seek time (cont’d) Reducing rotational latency RAIDs.

Similar presentations


Presentation on theme: "1 Storage (cont’d) Disk scheduling Reducing seek time (cont’d) Reducing rotational latency RAIDs."— Presentation transcript:

1 1 Storage (cont’d) Disk scheduling Reducing seek time (cont’d) Reducing rotational latency RAIDs

2 2 Reducing seek time (cont’d) Optimal algorithm? Requests arriving dynamically Length of queue may be short Total seek time is not the only criteria Fairness; Effect of file allocation Contiguous/non-contiguous Effect of Index blocks & directories Caching them in memory

3 3 Reducing rotational latency Hard for OS Disks do not disclose physical locations of logical blocks Disk scheduling algorithm in hardware (disk scheduler) But, OS may have other goals (different priority in I/O requests, order of disk writes…)

4 4 RAIDs Past: Redundant Array of Inexpensive Disks Now: Redundant Array of Independent Disks Hookup several disks to a single controller SCSI controller, IDE controller Special RAIDs controller From the user point of view A single large disk OS/Hardware hides RAID completely

5 5 Tradeoffs of using disk array Striping: read/write from multiple disks simultaneously Higher data transfer rate More prone to error Disk failure probability: p → one in n disk failure probability: np File spread out: one disk failure, entire file useless Add redundancy/error correction in disks Better reliability

6 6 RAID levels RAID0, RAID1, RAID2, RAID3, RAID4, RAID5 Different ways of adding redundancy & striping In practice, Many more variations have evolved Different implementation from different manufactures

7 7 Parity Checking Single Bit Parity: Detect single bit errors Two Dimensional Bit Parity: Detect and correct single bit errors 0 0

8 8 RAID 0 Striping at block level No redundancy Strip 0 Strip 4 Strip 8 Strip 1 Strip 5 Strip 9 Strip 2 Strip 6 Strip 10 Strip 3 Strip 7 Strip 11 Controller

9 9 RAID 1 Striping at block level Redundancy: mirrored disk Expensive: twice as many disks Strip 0 Strip 4 Strip 8 Strip 1 Strip 5 Strip 9 Strip 2 Strip 6 Strip 10 Strip 3 Strip 7 Strip 11 Controller Strip 0 Strip 4 Strip 8 Strip 1 Strip 5 Strip 9 Strip 2 Strip 6 Strip 10 Strip 3 Strip 7 Strip 11

10 10 RAID 2 Bit-level striping Redundancy: parity bits for error correction Not really used in practice parity disks

11 11 RAID 3 (bit-interleaved parity) Similar to RAID2 1-bit parity corrects 1-bit error Disk controller knows whether a sector is read correctly parity disk

12 12 RAID 4 (block-interleaved parity) Block-level striping Parity disk: parity block A block-access is slower than RAID3 Good for multiple read, large read/write Read-modify-write cycle parity disk

13 13 RAID 5 (block-interleaved distributed parity) Block-level striping Parity blocks spread over the disks, why? Most common parity RAID system Strip 0 Strip 4 Strip 8 Strip 1 Strip 5 Strip 9 Strip 2 Strip 6 Parity 2 Strip 3 Parity 1 Strip 10 Controller Strip 7 Strip 11 Parity 0

14 14 RAID summary RAID levels Different ways of redundancy No redundancy Parity (where to store parity) Error correcting codes Different ways of striping Bit level Block level


Download ppt "1 Storage (cont’d) Disk scheduling Reducing seek time (cont’d) Reducing rotational latency RAIDs."

Similar presentations


Ads by Google