RAID 1 read performance / myth or fact?

Toggle sidebar Toggle sidebar
S

spikespiegal

Golden Member
Oct 10, 2005
1,219
9
76
Help me settle an argument please.........

I've been working with Server based RAID controllers for a long time with pretty much all RAID levels.

I keep hearing that RAID 1 is supposed to increase read performance because data is duplexed on both drives. However, of all the zillions of RAID controllers I've worked with I've never seen a RAID 1 card improve read performance. They all impair write performance to a differening degree (depending on the quality of the controller) with read performance usually staying close to single drive levels.

Matter of fact, I've never seen a benchmark showing a RAID 1 card having improved RAID 1 performance over a single drive, but I keep reading that "if you have a good card, RAID 1 can improve read performance". I guess my 3Ware, Adaptec, Dell PERC, LSI and HP/Compaq Di controllers must be junk then. Maybe I should upgrade to Promise.....

If you know of such a benchmark showing improved Rad 1 read performance, could you URL it please?

Until then, I consider "this myth is busted".
 
P

PurdueRy

Lifer
Nov 12, 2004
13,837
4
0
Originally posted by: spikespiegal
Help me settle an argument please.........

I've been working with Server based RAID controllers for a long time with pretty much all RAID levels.

I keep hearing that RAID 1 is supposed to increase read performance because data is duplexed on both drives. However, of all the zillions of RAID controllers I've worked with I've never seen a RAID 1 card improve read performance. They all impair write performance to a differening degree (depending on the quality of the controller) with read performance usually staying close to single drive levels.

Matter of fact, I've never seen a benchmark showing a RAID 1 card having improved RAID 1 performance over a single drive, but I keep reading that "if you have a good card, RAID 1 can improve read performance". I guess my 3Ware, Adaptec, Dell PERC, LSI and HP/Compaq Di controllers must be junk then. Maybe I should upgrade to Promise.....

If you know of such a benchmark showing improved Rad 1 read performance, could you URL it please?

Until then, I consider "this myth is busted".
Click to expand...

Do you mean RAID 0? Raid 1 is pretty much only for backup purposes and there shouldn't per any performance difference associated with it.
 
Fullmetal Chocobo

Fullmetal Chocobo

Moderator<br>Distributed Computing
Moderator
May 13, 2003
13,704
7
81
Originally posted by: PurdueRy
Originally posted by: spikespiegal
Help me settle an argument please.........

I've been working with Server based RAID controllers for a long time with pretty much all RAID levels.

I keep hearing that RAID 1 is supposed to increase read performance because data is duplexed on both drives. However, of all the zillions of RAID controllers I've worked with I've never seen a RAID 1 card improve read performance. They all impair write performance to a differening degree (depending on the quality of the controller) with read performance usually staying close to single drive levels.

Matter of fact, I've never seen a benchmark showing a RAID 1 card having improved RAID 1 performance over a single drive, but I keep reading that "if you have a good card, RAID 1 can improve read performance". I guess my 3Ware, Adaptec, Dell PERC, LSI and HP/Compaq Di controllers must be junk then. Maybe I should upgrade to Promise.....

If you know of such a benchmark showing improved Rad 1 read performance, could you URL it please?

Until then, I consider "this myth is busted".
Click to expand...

Do you mean RAID 0? Raid 1 is pretty much only for backup purposes and there shouldn't per any performance difference associated with it.
Click to expand...

He is referring to RAID 1. The theory that he is speaking of is that the read performance of the array will be better than a single drive due to the fact that the controller is reading data from two sources instead of one, choosing the fastest route and increasing read speeds. I don't work in server platform environments or anything, but I've never noticed an increase in read speeds in real world or benchmark tests on any controller I've used (Promise, Highpoint, Areca, nor SATA or IDE).
 
Modelworks

Modelworks

Lifer
Feb 22, 2007
16,240
7
76
I worked with moving tons of data around on raid arrays and have never heard of raid 1 being used in any situations where speed was of concern.

When we worked with large video files, HD content, we always used raid 1+0.
Its not something for the home user though, expensive to set up.
 
J

jb1677

Junior Member
Jan 29, 2005
22
0
0
Originally posted by: spikespiegal
Help me settle an argument please.........

I've been working with Server based RAID controllers for a long time with pretty much all RAID levels.

I keep hearing that RAID 1 is supposed to increase read performance because data is duplexed on both drives. However, of all the zillions of RAID controllers I've worked with I've never seen a RAID 1 card improve read performance. They all impair write performance to a differening degree (depending on the quality of the controller) with read performance usually staying close to single drive levels.

Matter of fact, I've never seen a benchmark showing a RAID 1 card having improved RAID 1 performance over a single drive, but I keep reading that "if you have a good card, RAID 1 can improve read performance". I guess my 3Ware, Adaptec, Dell PERC, LSI and HP/Compaq Di controllers must be junk then. Maybe I should upgrade to Promise.....

If you know of such a benchmark showing improved Rad 1 read performance, could you URL it please?

Until then, I consider "this myth is busted".
Click to expand...

Some array controllers are capable of reading from both disks in a classic 2 drive RAID 1, I have never seen a low end controller (including IDE, SATA or SCSI) do this, they all tend to read from only one disk. Hp SmartArray controllers advertise this feature but I have never actually benchmarked it .... may have to set up something and test. I have never seen this feature advertised on any consumer grade controllers.

Even more expotic configurations can be accomplished by duplexing controllers, where you mirror a pair of controllers.

Jason

 
M

Madwand1

Diamond Member
Jan 23, 2006
3,309
0
76
My understanding is that many implementations don't do RAID 1 read performance enhancement at all. Some implementations do, but only for multiple concurrent reads. Therefore most single-user/single-transfer type tests won't show any improvement. And as most multi-user tests are heavily skewed towards random access performance, a significant improvement might not be easily seen there either.
 
E

erwos

Diamond Member
Apr 7, 2005
4,778
0
76
Originally posted by: Madwand1
My understanding is that many implementations don't do RAID 1 read performance enhancement at all. Some implementations do, but only for multiple concurrent reads. Therefore most single-user/single-transfer type tests won't show any improvement. And as most multi-user tests are heavily skewed towards random access performance, a significant improvement might not be easily seen there either.
Click to expand...
Right on the money. RAID 1 will theoretically give you a boost, but only under the right circumstances - rather like RAID 0, I should point out.
 
B

Blain

Lifer
Oct 9, 1999
23,643
3
81
Originally posted by: spikespiegal
Help me settle an argument please.........

I've been working with Server based RAID controllers for a long time with pretty much all RAID levels.

I keep hearing that RAID 1 is supposed to increase read performance because data is duplexed on both drives. However, of all the zillions of RAID controllers I've worked with I've never seen a RAID 1 card improve read performance. They all impair write performance to a differening degree (depending on the quality of the controller) with read performance usually staying close to single drive levels.

Matter of fact, I've never seen a benchmark showing a RAID 1 card having improved RAID 1 performance over a single drive, but I keep reading that "if you have a good card, RAID 1 can improve read performance". I guess my 3Ware, Adaptec, Dell PERC, LSI and HP/Compaq Di controllers must be junk then. Maybe I should upgrade to Promise.....

If you know of such a benchmark showing improved Rad 1 read performance, could you URL it please?

Until then, I consider "this myth is busted".
Click to expand...
* Are you asking about SCSI, SATA or IDE HDs?
* You're only interested in hardware based controllers, correct?
* What chunk-size are you using?

 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom