Memory Raid?

puuff · 2011-03-09 21:54:55

I have no technical insights when it comes to memory technologies but one thought have crossed my mind. What if you created some kind of RAID configuration for two (or more) memory modules? Could that (in theory) give you a better read and/or write performance? Maybe it's just a bad idea as e.g RAID-0 gives better read but worse write performance and the same (but opposite) goes for RAID-1.

Edit: Shouldn't write threads when I haven't slept enough. I totally forgot about memory dual channels xd. This is on the house xd.

Last edited by puuff (2011-03-09 22:13:18)

IncredibleLaser · 2011-03-10 20:36:15

RAID-0 gives better read but worse write performance and the same (but opposite) goes for RAID-1.

No. RAID 0 is double the speed of a single harddisk, both read and write and offers the size of the smaller disk x 2, but the danger of losing data is higher.

fukawi2 · 2011-03-10 22:51:32

IncredibleLaser wrote:

...but the danger of losing data is higher.

Double, in-fact

student975 · 2011-03-10 23:09:23

fukawi2 wrote:

Double, in-fact

Why?

Say, an opportunity to fail for an HDD during given period is 0.1. For RAID0 an opportunity of "all is OK" is 0.81, i.e. we have an opportinity of lost data equal to 0.19. For RAID1 we have an opportunity of lost data equal to 0.01.

Am I wrong?

TigTex · 2011-03-11 00:24:47

"memory modules" = RAM memory?
the "raid" it's called dual-channel and triple-channel on them

with raid0, if one hdd fails, all your data is lost from both hdd's. So... if you have 2 hdd's, there's twice the possibility of data loss.

wrong section to post this

fukawi2 · 2011-03-11 05:01:27

student975 wrote:

fukawi2 wrote:
Double, in-fact
Why?

TigTex wrote:

with raid0, if one hdd fails, all your data is lost from both hdd's. So... if you have 2 hdd's, there's twice the possibility of data loss.

So it's actually:

RAID-0 Data Loss Risk = N x Risk of Single Hard Drive Failure

Where N = Number of Drives in the Array.

IncredibleLaser · 2011-03-11 09:41:34

No. Say you have 2 hard drives, each with a possibility of 0.5 to fail within a given period of time, your chance that your RAID still works is not 0, but 0.25. As student975 said, it's the product of the possibility of survival of each hdd in your RAID.

fukawi2 · 2011-03-11 12:41:12

No, a failure of *any* hard drive in a RAID-0 array destroys the whole array (remember the data is striped).

Any form of RAID increases the chance of experiencing disk failure. The level of RAID determines the impact of the failure on your data.

IncredibleLaser · 2011-03-11 23:20:21

fukawi2 wrote:

No, a failure of *any* hard drive in a RAID-0 array destroys the whole array (remember the data is striped).

I do know that. But your math is flawed. When I wrote product, I meant that in a mathematical way.

fukawi2 · 2011-03-12 00:32:43

IncredibleLaser wrote:

I do know that. But your math is flawed.

Oh... Well yeah, maths isn't my best skill
Apologies for the misunderstanding

TomB17 · 2011-03-15 17:01:56

I'd just like to point out that parity bits and ECC are not dissimilar to the technology used in RAID. They address reliability issues, not speed issues.

TigTex already addressed the speed issue with the bus width discussion.

Arch Linux

#1 2011-03-09 21:54:55

Memory Raid?

#2 2011-03-10 20:36:15

Re: Memory Raid?

#3 2011-03-10 22:51:32

Re: Memory Raid?

#4 2011-03-10 23:09:23

Re: Memory Raid?

#5 2011-03-11 00:24:47

Re: Memory Raid?

#6 2011-03-11 05:01:27

Re: Memory Raid?

#7 2011-03-11 09:41:34

Re: Memory Raid?

#8 2011-03-11 12:41:12

Re: Memory Raid?

#9 2011-03-11 23:20:21

Re: Memory Raid?

#10 2011-03-12 00:32:43

Re: Memory Raid?

#11 2011-03-15 17:01:56

Re: Memory Raid?

Board footer