Midrange Performance Fundamentals Flashcards

1
Q

What is Cache Dirty Size?

A

capacity of mirrored cache holding data yet to be committed to drive

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is a Vault?

A

internal area within the storage system that is required for write cache dumping

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What does it mean to coalesce IO?

A

to combine multiple smaller IO into a larger IO

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is locality?

A

active data regions relative to logical block address range or aggregate of ranges

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What parts of RAID affect performance the most?

A

data copies incur additional drive IO (RAID 1/0)
parity generation will take CPU utilization (RAID 5 and 6)
parity operations will take drive utilization (RAID 5 and 6)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Why are SSDs capable of much higher IOPS?

A

multiple internal channels allow multiple operations in parrallel

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What can affect the performance on an SSD?

A

wear leveling
garbage collection
endurance levels

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is an example of availability?

A

10 drives have higher performance potential than 5 drives but a lower availability due to the higher probability of multiple drive failures

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What are the ways to choose a RAID type?

A

fast - for a given number of spindles how good is the relative performance

cheap - how much usable capacity is there - compared to the raw capacity

available - how unlikely is it that the next drive failure will cause data loss

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What is the write penalty for each RAID group?

A

RAID 5 - 4 - old/new data, old/new parity 1
RAID 6 - 6 - old/new data, old/new parity 1, old/new parity 2
RAID 1/0 - 2 - old/new data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What causes more drive load on parity RAID types?

A

random writes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is almost always true about midrange storage and performance?

A

random reads almost always a cache miss

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What does a smaller IO help with?

A

transactional capability and faster service times

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is the benefit of larger IO?

A

affords higher bandwidth - but large random can result in additional drive IO if spanning multiple stripe elements

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What does random IO benefit from?

A

lots of drives

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is the relationship between concurrency and storage consolidation?

A

traditionally leads to more concurrency (multiple hosts pointing to one storage system)

17
Q

Why does server virtualization typically reduce concurrency?

A

consolidating servers means consolidating paths to the storage

18
Q

What benefit does read caching have on performance?

A

previously read data is stored in cache - LRU data is discarded when cache pages are needed

much shorter response time to service a cached read

19
Q

What benefit does write caching have on performance?

A

hitting write cache typically faster than going to HDD (as long as cache isn’t saturated)

for a small IO under 1ms vs 5-10ms going to disk

insulates host IO from drive response (RAID overhead, drive type, etc.)

can coalesce chunks of data up to entire stripes

20
Q

What may not benefit from write cache?

A

a storage system that has a lot of non-parity writes and uses SSDs

21
Q

What benefit does IO ordering/processing have on performance?

A

while data is stored in cache controller can optimize how it will be sent to the drives

order blocks to minimize drive seeking during reads and writes

coalesces multiple contiguous blocks into a single large IO