Fault Tolerance

Question 1

What is dependability?

Answer 1

can we expect the delivered service to match the specified service?

Question 2

What is a fault?

Answer 2

Module deviates from the specified behavior. Ex. a function that works except 2+5=8

Question 3

What is an error?

Answer 3

Actual behavior in the system differs from specified behavior (an instance of a fault). Ex. We call add(5,3) and store 7 in a register

Question 4

What is a failure?

Answer 4

The system deviates from specified behavior (when an error causes our program to do something wrong). Ex. Schedule a meeting for 7am instead of 8.

Question 5

What is MTTF?

Answer 5

Mean time to failure. average time of continuous service accomplishment before a failure

Question 6

What is MTTR?

Answer 6

average time to repair

Question 7

What is a hardware fault?

Answer 7

hardware fails to perform as designed

Question 8

What is a design fault?

Answer 8

software bugs, hardware design mistakes, FDIV bug - usually permanent

Question 9

What is an operational fault?

Answer 9

operator, user mistakes

Question 10

What is an environmental fault?

Answer 10

fire, power failure, sabotage etc.

Question 11

What is a permanent fault?

Answer 11

One that is permanent

Question 12

What is an intermittent fault?

Answer 12

lasts for a while, but recurring. Ex. Overclocking - works fine for a while, then crashes, works fine for a while once rebooted

Question 13

What is a transient fault?

Answer 13

lasts for a while, but then goes away

Question 14

What is fault avoidance?

Answer 14

prevent faults from occurring

Question 15

What is fault tolerance?

Answer 15

prevent faults from becoming failures. Ex. Redundancy

Question 16

What are some fault tolerance techniques?

Answer 16

Checkpoint (recovery), Dual and Triple Module Recovery (DMR, TMR)

Question 17

What is N-Way Redundancy?

Answer 17

N modules do the same work, then vote on the answer. Can detect and recover from N-2 faults.

Question 18

What are some fault tolerance techniques for memory and storage?

Answer 18

Error detection/correction codes, RAID

Question 19

What is bit parity?

Answer 19

one extra bit added to the end of data (XOR of all bits). Can detect single-bit errors

Question 20

What is ECC?

Answer 20

Error-correction code. Method to detect and fix errors

Question 21

What is RAID?

Answer 21

Redundant Array of Independent Disks. Several disks playing the role of one disk

Question 22

What is RAID0?

Answer 22

“Striping”. Take one disk and put all the odd tracks on one disk and all the even tracks on another

Question 23

What is RAID1?

Answer 23

“Mirroring”. Same data on both disks. Can correct any fault that affects 1 disk.

Question 24

What is RAID4?

Answer 24

N disks, (N-1) disks are stripped. Nth disk is parity disk.

Question 25

What's the drawback to RAID4?

Answer 25

parity disk becomes the bottleneck

Question 26

What is RAID5?

Answer 26

Distributed block-interleaved parity. Rotate the parity block across all disks.

Question 27

What is RAID6?

Answer 27

Have two parity blocks for each set of data blocks. Each parity block is computed differently. Can recover from 2 failures.

Question 28

What is RAID6?

Answer 28

Have two parity blocks for each set of data blocks. Each parity block is computed differently. Can recover from 2 failures.