How good is ride 1 for speeding up random disk reads?

Question

Iron

Who has been? Are there any pitfalls and in general does it make sense to increase the speed of random reading with the help of ride 1? What are the pros and cons?
Sorry for the slightly vague question.

Answer 1

Ai-yay-yay, we don’t know how to google (

Answer 2

How good is ride 1 for speeding up random disk reads?

At 0.9-1 as a rule.
The speed should decrease slightly, or remain the same as that of one disk.
It is extremely rare for normal mirror controllers to increase the speed. about a multiple of the number of disks.
But to talk about how good RAID is for increasing the speed of random reading when there is an SSD - it makes no sense.
And given that you can already buy Intel Optane on sale - in general, the question seems strange.

Answer 3

the speed of ride0 will double (approximately).
data placement on raid0 will be identical on both disks, therefore, the time of random jumping of disk heads will not change (i.e. the speed will be the same twice as high)

Answer 4

RAID 1 - no way
RAID 0 - reduces reliability
Random read speed well increases SSD

Answer 5

If you have a good raid controller, then both Raid0 and Raid1 will increase random reading approximately the same - by the number of devices in the raid.
If a built-in raid on a cheap motherboard, especially in which joyet0 is written instead of Raid0, then you may not notice anything.
With software I will not tell. In theory, the speed should increase, but by how much ...

Answer 6

There are options for assembling RAID-1, or rather controllers, when the read speed is increased. Not at the RAID-0 level, but it will be noticeable.
Parallelized reading can be done by advanced controllers. Soft hardly.

Answer 7

You have RAID 1 in the title, and RAID 0 in the text of the question.
Judging by the fact that the emphasis is on speed, apparently, we are talking about the latter.
This is not very relevant for us, so I didn’t try to keep the information in my head, but if my memory serves me, it makes sense to reduce the size of the stripe to increase the throughput, but if the priority is random I / O, then vice versa, increase it.
At the same time, it is important to remember such a parameter as queue depth. If it is equal to one all the time, then it is clear that there will be no increase, because. Only one request is sent per unit of time.
Of course, there will be no infinite linear growth, but I think that with QD = 16 / 32 / 64 you will be close to the possible maximum.
Also, do not forget about the implementation of reading in drivers. If the manufacturer changes or optimizes something for specific operating systems, then this will affect the speed.
In general, experiment, but do not forget that RAID 0 has low reliability, which decreases with the increase in the number of disks. And if you lose one drive, you lose access to the entire array.

Answer 8

Random reading is a very vague term, it all depends on the type of data, and of course fs, if we are talking about the database, then it is better to start with raid 10 and optimize the database at the installation stage
mat part
www.ibase.ru/raid

Answer 9

If this is a workstation, then put raid0 + permanent backup of important files, if it's a server, then it's not worth it.
In general, there is an increase in speed, but insignificant, let's say, it's cheaper and more profitable to install an ssd.

Answer 10

I tested soft raid1 on linux and windows 7, on Linux the reading speed did not increase, but on windows it unexpectedly doubled, so the soft raid on windwos turned out to be pretty good.