2TB HDD to use in RAID5 on Intel ICH10R

[benzebut]

I've done the mistake to buy some WD Caviar Green drives for RAID5.
On Intel ICH10R onboad raid controller it gives poor performance. Found out its due to a disabled feature on my green drive as well as much of today's drives.

The feature allows improved error handling in a RAID environment. In some cases, there is a conflict as to whether error handling should be undertaken by the hard drive or by the RAID controller, which leads to drives being marked as unusable and significant performance degradation, when this could otherwise have been avoided.

Each vendor use their own version of it.

Western Digital uses TLER
Seagate uses ERC
Samsung and Hitachi use CCTL
http://en.wikipedia.org/wiki/Time-Limited_Error_Recovery

I am trying to figure out which drive have this feature or not.
Here's a brief list of 2TB drives

Brand - Series - TLER,ERC or CCTL
Hitachi - Deskstar 5K3000 - CCTL 50ms
Hitachi - Deskstar 7K2000 - CCTL 50ms
Hitachi - Deskstar 7K3000 - CCTL 50ms
Hitachi - Ultrastar 5K3000 - CCTL 50ms
Hitachi - Ultrastar 7K3000 - CCTL 50ms
Hitachi - Ultrastar A7K2000 - CCTL 50ms
Samsung - EcoGreen F4 - No
Seagate - Barracuda Green - No
Seagate - Barracuda XT - Can't find
Seagate - Constellation ES - Can't find
Western Digital - Caviar Green - No
Western Digital - Caviar Black - No
Western Digital - RE4 - TLER
Western Digital - RE4-GP - TLER

This is a first draft, there might be some model/firmware offering different capabilities. Any help to fill it out would be appreciated.

 

[FishAk]

Search this forum for; "Caviar Blacks in RAID without TLER OK"

 

[benzebut]

Search this forum for; "Caviar Blacks in RAID without TLER OK"

Business Critical RAID Environments – WD Caviar Black Hard Drives are not recommended for and are not warranted for use in RAID environments utilizing Enterprise HBAs and/or
expanders and in multi-bay chassis, as they are not designed for, nor tested in, these specific types of RAID applications. For all Business Critical RAID applications, please consider WD’s
Enterprise Hard Drives that are specifically designed with RAID-specific, time-limited error recovery (TLER), are tested extensively in 24x7 RAID applications, and include features like​

enhanced RAFF technology and thermal extended burn-in testing.

http://www.wdc.com/wdproducts/library/SpecSheet/ENG/2879-701276.pdf

 

[8steve8]

i dont think the lack of tler is responsible for your low write performance, what caching do you have enabled?

btw i currently have a raid5 array of "green" drives. no problem with perfrmance.

also there is a utility to turn on tler on any WD drive.

 

[FishAk]

It's understood by most that manufactures would like to get the most money possible for their products. One way to get more money is by changing the firmware on otherwise identical products to exclude a certain usage set. If you are willing to pay more money, than by all means, buy the more expensive disks. If you would like to save money, than you should learn how to use the product in your proposed manor, even while unsupported by the manufacturer.

I get the impression you didn't read through the thread I directed you to. In it, I explained how I was able to get around the problem of not being able to limit error recovery.

Although like WD, Samsung also would like you to buy their enterprise drive, there are much fewer reports of the Samsung disks dropping out of arrays. Because they are also cheaper and perform better, I recommend them. However, if you want any of the manufacturers to stand behind your array, and say they support it, you will have to pay extra to get their respective firmware.

That said, the performance issue you have isn't affected by the error recovery. If the disks have enough errors that they need to recover from- enough to effect the performance- the disks are failing. SW RAID 5 “can” perform well, but you need to have all aspects lined up perfectly, including stripe size, alignment, allocation units, the number of disks used, ect.

Raid 10 is much more forgiving, and it also doesn't suffer from the “write hole” that can cause a RAID 5 array to be unrecoverable. Additionally, the RAID 10 array can recover much faster from a failed disk, providing less opportunity for a second failed disk to destroy all data on the array. On a 4 disk RAID 0+1 array, as provided by ICH10R, there is actually a 33 percent chance that a second disk failure will still result in the array still remaining functional.

also there is a utility to turn on tler on any WD drive.

There is a utility, but WD made it so the new disks (later than 11/2009) can't use it.

 

[benzebut]

i dont think the lack of tler is responsible for your low write performance, what caching do you have enabled?

btw i currently have a raid5 array of "green" drives. no problem with perfrmance.

also there is a utility to turn on tler on any WD drive.

Hi,

What is your setup for your raid5?
# of drives and brand/serie/model ?
What raid controller you using? onboard ICHxR?

Also, could you post a hdtune or something so i can compare to?
thx

 

[benzebut]

It's understood by most that manufactures would like to get the most money possible for their products. One way to get more money is by changing the firmware on otherwise identical products to exclude a certain usage set. If you are willing to pay more money, than by all means, buy the more expensive disks. If you would like to save money, than you should learn how to use the product in your proposed manor, even while unsupported by the manufacturer.

I get the impression you didn't read through the thread I directed you to. In it, I explained how I was able to get around the problem of not being able to limit error recovery.

Although like WD, Samsung also would like you to buy their enterprise drive, there are much fewer reports of the Samsung disks dropping out of arrays. Because they are also cheaper and perform better, I recommend them. However, if you want any of the manufacturers to stand behind your array, and say they support it, you will have to pay extra to get their respective firmware.

That said, the performance issue you have isn't affected by the error recovery. If the disks have enough errors that they need to recover from- enough to effect the performance- the disks are failing. SW RAID 5 “can” perform well, but you need to have all aspects lined up perfectly, including stripe size, alignment, allocation units, the number of disks used, ect.

Raid 10 is much more forgiving, and it also doesn't suffer from the “write hole” that can cause a RAID 5 array to be unrecoverable. Additionally, the RAID 10 array can recover much faster from a failed disk, providing less opportunity for a second failed disk to destroy all data on the array. On a 4 disk RAID 0+1 array, as provided by ICH10R, there is actually a 33 percent chance that a second disk failure will still result in the array still remaining functional.



There is a utility, but WD made it so the new disks (later than 11/2009) can't use it.

Thx for your concern,

Ive read your whole post "Caviar Blacks in RAID without TLER OK"
But im not quite sure i understand your final solution, is it to do a raid10 or raid0+1 array?

I ran a smart test on all my 3 drives and they are all fine. Their speeds are ok. How come when paired in a raid5 array, the speed drop so badly?

I'm using latest RST driver, the array is fully initialized and write-back cache is enabled. Is there something im missing? How can i troubleshoot that more further?

 

[FishAk]

First, you should understand that enabling write-back cache is very dangerous. If there is a hiccup in the kernel that causes a crash, or a power interruption to the mother board, the risk of corrupting the array to a non recoverable state is very real.

I played around with RAID 5 a little, but I was unable to achieve good performance, Coupled with the dangers, and the trivial savings in cheap disk space, I abandoned it. I do see reports of people getting good performance, but the stars have to be aligned just right.

There is no such thing as RAID 10. RAID 10 is shorthand for RAID 1 + 0, or RAID 0 + 1. RAID 0 + 1 is a mirrored set of striped disks. RAID 1 + 0 is a striped set of mirrored disks. The difference may sound academic, but the durability of the array has an advantage with RAID 1 + 0. This is because 1+0, has a 66 percent chance of the array surviving with two broken disks. With 0 + 1, the chance of survival after a second failure is only 33 percent. Unfortunately, ICH10R uses 0 + 1.

My solution to dropped disks was to create a RAID 0 array with a small portion of my 4 matching disks, and then make a RAID 10 array with the remaining space. I think that since there is a RAID 0 portion, the controller doesn't allow the disks to be dropped due to error recovery timeout, since this would obliterate the RAID 0 array (and make any error recovery mute).

 

[Emulex]

RE4 work great. stick to those. they are cheap for the amount of storage you get. but remember that is only one piece of the puzzle. power never goes on on raid-5. systems never freeze up on raid-5.

just buy one of my LSI raid controllers and do raid-10 with RE4's and you'll be good to go.