Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
37 Cards in this Set
- Front
- Back
What is RAID?
|
Redundant Arrays of Inexpensive Disks
|
|
What is data 'striping'?
|
Distribution of data accross multiple disks
|
|
What are the benefits of striping?
|
Maximises data transfer rate on large accesses, allows multiple IO requests in parallel, reducing queueing time
|
|
What is the risk of RAID?
|
Heightened risk of failure
|
|
Why do we use redundancy?
|
To cope with failure
|
|
How can we improve the reliability of a system?
|
NOT by RAID. We can use fewer components or improve manufacturing technology
|
|
What is a problem with redundancy?
|
It is expensive - write operations must update data in all redundant disks
|
|
Summarise the goals of redundancy.
|
To maximise the number of disks accessed in parallel, minimise amount of disk space for redundancy and minimise the overhead of both.
|
|
Describe the concept of Fine-Grained arrays
|
Interleaving of data in small units so all IO requests access all disks
|
|
Pro and Con of fine-grained?
|
High data transfer rate for IO but only one IO servicable at once (as all disks used for each operation)
|
|
Describe Coarse-grained arrays.
|
Data is interleaved in large units, such that small IO accesses small number of disks while large IO still can access all disks.
|
|
What is the benefit of using coarse-grained interleaving?
|
Multiple small requests can be serviced at once, and still get high transfer rate for large IO
|
|
What are the issues in question when determining redundancy?
|
Which methods to use to compute redundant information and how to distribute across the array.
|
|
What is a benefit of uniform distribution of redundant data?
|
Avoid 'hot spots' where we can get issues with load balance.
|
|
What is RAID-0?
|
No redundancy.
|
|
Why doesn't RAID-0 give better performance?
|
Arrays of REDUNDANT data allows optimal selection for fastest data retrieval.
|
|
When might RAID-0 be used?
|
When performance and capacity are key, such as in supercomputing
|
|
What is RAID-1?
|
Mirroring.
|
|
How many disks does RAID-1 use compared with RAID-0?
|
Twice as many - each disk has a mirror copy
|
|
How is performance optimised in RAID-1?
|
Which disk is read depends on delays
|
|
When might RAID-1 be used?
|
When availability is key, such as databases
|
|
How does RAID-1 perform with IO requests?
|
Up to twice as many requests can be made as for RAID-0
|
|
What is RAID-2?
|
Parallel access - redundancy through Hamming code
|
|
How many disks are accessed for a single write in RAID-1?
|
All!
|
|
What is the proportionality between number of redundant disks in RAID-1 and total number of disks?
|
Log to base 2, so 32 disks : 5 redundant (2^5 = 32)
|
|
What does this mean?
|
Storage efficiency increases with more data disks.
|
|
What is RAID-3?
|
Bit-Interleaved Parity; where data is reconstructed from a 'parity disk'
|
|
How is parity calculated for an array of five disks?
|
D1(k) = D4(k) xor D3(k) xor D2(k) xor D0(k)
|
|
Can the parity disk participate in reads?
|
No; it contains no data, just parity information.
|
|
Can the parity disk participate in writes?
|
Yes, it must to update info
|
|
What is a disadvantage of using RAID-3?
|
The bottleneck on the parity disk
|
|
What level or RAID do we not look at on this course?
|
RAID-4
|
|
What is RAID-5?
|
Block-interleaved distributed parity; Similar to RAID-3 though parity is distributed over all data disks
|
|
What is RAID-5 best at?
|
Small reads and large writes.
|
|
What is RAID-5's biggest weakness?
|
Small writes - Every write requires modification of the parity so is inefficient compared with mirroring.
|
|
What is RAID-6?
|
P+Q Redundancy - uses two different parity calculations on separate disks.
|
|
What is RAID-6's greatest strength?
|
Data is fully recoverable from two disk failures!
|