Jump to content
MvL

5400 or 7200 HDD

33 posts in this topic Last Reply

Recommended Posts

Hi,

 

I think you people have more experience as me with unRAID servers. I want to add a 2TB parity drive to my test unRAID box. I thought that i read to use a 7200RPM HDD as parity drive. Is this right? What for drive do you recommend as data drive? The Green power models 5400RPM or just a 7200RPM model.

Share this post


Link to post

Correct.

 

In my experience, a 7200rpm drive acting as parity will provide slightly increased performance over a 5400/5900rpm "Green" drive. I currently use a 2TB WD Black (model: WD2001FASS) as my parity drive. I have a few Green drives (1.5TB WD EARS) in my system.

 

Share this post


Link to post

From all my reading (done a whole bunch), you want the parity drive to be the largest (since it'll restrict how much is actually used by the data drives) and you want it to be the fastest.

 

The reason for this is any writes would be as fast as your slowest drive being used (ie...data & parity). And since you can be writing to multiple data drives and they would also be needing to write to the parity drive, you could have writes being delayed if you end up with an IO queue on the parity drive.

 

 

 

Share this post


Link to post

If all your disks are 5400 RPM and you will only be writing to one disk at a time (or using a cache drive), then a 7200 RPM parity drive will not add much to performance.

 

If you have mixed RPM drives and you write to multiple drives simultaneously then the fastest parity drive you can find will provide a slight benefit.

 

If you have a cache drive, that benefit will not be seen.

 

If you are writing directly to the array with some internal program (news downloader, torrent, transcoder) then you will see a benefit.

 

For average media storage use, the cache drive provides a big benefit.

For internal applications writing to the array drives, the fastest parity drive you can find/afford provides a big benefit.

 

It is true, your array will function as fast as your slowest drive.

Yet if you use multiple drives, a faster parity drive helps.

Share this post


Link to post

Hi,

 

Thanks for the answers.

 

So if you have a cache drive then i make not a lot of difference if you use a 5400/5900/7200 drive. I'm planning to use a cache (when i buy my licence) drive and i'm very curious what the write speeds are when you use SSD all-thought they are expensive.

Share this post


Link to post

Hi,

 

Thanks for the answers.

 

So if you have a cache drive then i make not a lot of difference if you use a 5400/5900/7200 drive. I'm planning to use a cache (when i buy my licence) drive and i'm very curious what the write speeds are when you use SSD all-thought they are expensive.

 

I don't think it's worth the expense for the average user.

The cost could be put into a 2TB drive which could become a replacement drive in an emergency.

(And you would still save money).

Share this post


Link to post

I agree.  I use an older 320 GB Seagate 7200 rpm w/ 8 mb cache as my cache drive.  I routinely see write speeds of around 65-70 mb/s.  In my tests with two SSDs (one 30 GB OCZ Agility and one 60 GB Corsair), the max transfer speed I saw was 73 mb/s.  Not much of a difference, and definitely not worth the difference in price.  My advice is to take your oldest, slowest, and smallest drive and make that your cache drive.  Or if you want to have a warm spare on hand, use a 2 TB green drive as WeeboTech suggested.

Share this post


Link to post

I use a fairly fast RAID1 driveset for my cache, just not an SSD. (yet)

 

For my particular use I may upgrade to a SSD, but this is only because I set up special hidden directories and use NFS & SAMBA to provide a single home directory for all my network logins.

 

In this case, SSD is great as I do not have to wait for spin up.

 

But for the average unRAID use, a regular drive as large as your largest drive is what I would recommend.

One that is "warm spare" capable.

Share this post


Link to post

Hi,

 

I use a fairly fast RAID1 driveset for my cache, just not an SSD. (yet)

 

Just out of curiosity is this hardware raid 1 or sofware raid 1(mdadm). You could set two drives in raid 0 to get extra performance but this is not safe i suppose..

 

But for the average unRAID use, a regular drive as large as your largest drive is what I would recommend.

One that is "warm spare" capable.

 

What do you mean with warm spare capable?

 

 

 

 

Share this post


Link to post

Meaning that if one of your data or parity drives fail, you can use the cache drive as a quick replacement for the failed drive.

 

Hence the warm spare...

Share this post


Link to post

Oh, Thanks for explaining!

 

Meaning that if one of your data or parity drives fail, you can use the cache drive as a quick replacement for the failed drive.

 

Hence the warm spare...

Share this post


Link to post

I use a fairly fast RAID1 driveset for my cache, just not an SSD. (yet)

 

Just out of curiosity is this hardware raid 1 or sofware raid 1(mdadm). You could set two drives in raid 0 to get extra performance but this is not safe i suppose..

 

See these posts

http://lime-technology.com/forum/index.php?topic=7150.0

http://lime-technology.com/forum/index.php?topic=2113.0

 

I do RAID0 Parity, RAID1 Cache.

 

You cannot use mdadm software raid as the unRAID subsystem replaces the normal linux kernel md raid subsystem.

So I use the Areca ARC-1200 for hardware raid. You can find some good deals in auctions, you just have to wait.

But the return on investment may not be worth it for the average use.

 

Since I use my unRAID server for daily file services, I need it to be as fast as possible.

 

This weekend I did some adjustments, tunings and I got over 45MB/s.

Then I adjusted the kernel and how it caches the data and saw it peak at 70MB/s.

For me this is good as a burst under a certain size moved as fast as possible.

It slowed down much more with Tera copy and a long arduous copy task, but bursting at 70MB/s made my day!

 

This is because I use NFS to mount the drives remotely for my source code.

So I need it to read/write as fast as possible when compiling.

Share this post


Link to post

interesting.. I would be awesome when you can write with 70MB/sec or maybe 100MB/sec. Hmm, you also need a fast network cat6 or cat 6a. Thanks for explaining.

Share this post


Link to post

interesting.. I would be awesome when you can write with 70MB/sec or maybe 100MB/sec. Hmm, you also need a fast network cat6 or cat 6a. Thanks for explaining.

 

The 70MB/s was not over the network. It was a local dd test and it was a burst for 256MB/s. it slowly dropped to about 65MB/s after 1GB or so.

For me that was fine. I need to read and update many small files fast in order to compile projects with a decent speed.

 

I do not believe 100MB/s is reachable. The drive alone does 110 to 120MB/s.

Share this post


Link to post

Hopefully I am not changing the direction of this topic...but I am also curious about 5400 vs 7200 rpm drives.  Currently I am running 2 x 1TB WD Caviar Black drives, and this weekend I added 2 x 250GB Caviar drives ("old" 7200 rpm drives).  While performing a parity check I took notice that the read speed of the actual check was slower than with the Caviar Black drives alone.  From my understanding, writes are only as fast as the slowest drive in the system - is this also the case during parity checks (e.g. reads are only as fast as the slowest drive)?

Share this post


Link to post

is this also the case during parity checks (e.g. reads are only as fast as the slowest drive)?

 

Yes , but only for the duration of the slower drive's size.

After that the parity check speeds up.

 

Keep in mind that every drive you add to the system slows it down.

If the parity check speed is in the mid 60 to 80 range it's OK.

Total time to generate parity needs to be considered. if it takes 2 days to recreate/check parity then something needs to be enhanced/upgraded.

 

Share this post


Link to post

Thank you for the clarification WeeboTech - I certainly appreciate it.  With my Caviar Blacks I believe my parity check speed was in the 100 to 110 range, but has dropped to around 80.  However, this was during start up and I did not monitor for any length of time.  So I presume this was the result of checking the slower drives first.

 

From what you mention, do parity checks follow the process of: parity drive works with a single data drive to perform calculations >> parity drive completes the check for that drive >> parity drive moves onto the next data drive?

 

Another question if I may... If the above is correct, what drives does unRAID begin the parity checking process?  (e.g. highest sata channel - or highest unRAID assigned drive?)

 

PS - My apologies for my poor use of "terminology" here  :-\

Share this post


Link to post

Another question if I may... If the above is correct, what drives does unRAID begin the parity checking process?  (e.g. highest sata channel - or highest unRAID assigned drive?)

 

Early on in the parity check all drives will be used.  Late in the parity check only the largest drives will be used.  For example:

 

Let's say your server has a 2 TB parity drive, four 1 TB data drives, and one 500 GB data drive.  Forget the spindle speed for the time being.  For the first block of time all 6 drives will be read (and the parity drive will be written to if any errors are detected).  Once the entirety of the 500 GB drive has been read it will no longer factor into the parity check speed and it will spin down in 1 hour (or whatever the spindown timer is set to).  The first block of time will be the slowest portion of the parity check.  In the second block of time only the 2 TB parity drive and the four 1 TB data drives will be read.  This will be faster than the first block of time since less drives are involved.  Once the 1 TB drives are read in their entirety, they will spin down (again after an hour or however long).  The third and final block of the parity check will involve only the 2 TB parity drive.  This will be by far the fastest block of the parity check since only one drive is involved, but it will still take some time.  Once the third block completes the parity check will be complete.

 

Now let's throw some numbers in there.  Let's say that the 500 GB drive is a slow 5400 rpm drive, two of the 1 TB drives are 5400 rpm and the other two are 7200 rpm, and the parity drive is 5400 rpm.  The parity check could basically then be broken up into four blocks of relative speed:

1) All drives involved, slowest block

2) 500 GB drive complete, slow block

3) Both 1 TB 7200 rpm drives complete, both 1 TB 5400 rpm drives still going, fast-ish block

4) All data drives complete, fastest block

 

So in this example there will still be 5400 rpm drives involved as long as there are any data drives involved.  You can also see that swapping out the parity drive for a 7200 rpm drive would only speed up the 4th and final block, since every other block still involves a 5400 rpm data drive.

 

Play around with this logic and hopefully you'll be able to figure out where 7200 rpm drives can help and where they won't help.  There are cases where a 7200 rpm parity drive will increase speeds and other cases where it won't make any difference at all.

 

Wow, I hope that all makes sense.  It sounds a lot more complicated than it is when I write it all out like that...

Share this post


Link to post

Another question if I may... If the above is correct, what drives does unRAID begin the parity checking process?  (e.g. highest sata channel - or highest unRAID assigned drive?)

 

Early on in the parity check all drives will be used.  Late in the parity check only the largest drives will be used.  For example:

 

I write it all out like that...

 

So this is about parity check. The time is takes to parity check the system.

 

This has no consequence when using a cache drive to write to the unRAID server. What happens if you don't use a cache drive, what you explained?

 

 

interesting.. I would be awesome when you can write with 70MB/sec or maybe 100MB/sec. Hmm, you also need a fast network cat6 or cat 6a. Thanks for explaining.

 

The 70MB/s was not over the network. It was a local dd test and it was a burst for 256MB/s. it slowly dropped to about 65MB/s after 1GB or so.

For me that was fine. I need to read and update many small files fast in order to compile projects with a decent speed.

 

I do not believe 100MB/s is reachable. The drive alone does 110 to 120MB/s.

 

 

dd test? I have to google on that.

 

Share this post


Link to post

Another question if I may... If the above is correct, what drives does unRAID begin the parity checking process?  (e.g. highest sata channel - or highest unRAID assigned drive?)

 

Early on in the parity check all drives will be used.  Late in the parity check only the largest drives will be used.  For example:

 

I write it all out like that...

 

So this is about parity check. The time is takes to parity check the system.

 

This has no consequence when using a cache drive to write to the unRAID server. What happens if you don't use a cache drive, what you explained?

 

A cache drive doesn't factor into a parity check at all since it is outside the parity-protected array.  So using one or not using one won't change what I described above.

 

interesting.. I would be awesome when you can write with 70MB/sec or maybe 100MB/sec. Hmm, you also need a fast network cat6 or cat 6a. Thanks for explaining.

 

The 70MB/s was not over the network. It was a local dd test and it was a burst for 256MB/s. it slowly dropped to about 65MB/s after 1GB or so.

For me that was fine. I need to read and update many small files fast in order to compile projects with a decent speed.

 

I do not believe 100MB/s is reachable. The drive alone does 110 to 120MB/s.

 

 

dd test? I have to google on that.

 

 

A dd test is just the computer making up random data to read and write to a drive so that it can test the drive's maximum speeds.  The dd test gives you the maximum thresholds of your drive's capabilities.  Transferring real data will generally be a bit slower, depending on the type of data (large movie files vs a folder full of small images and documents, for example).

Share this post


Link to post

dd test? I have to google on that.

 

A dd test is just the computer making up random data to read and write to a drive so that it can test the drive's maximum speeds.  The dd test gives you the maximum thresholds of your drive's capabilities.  Transferring real data will generally be a bit slower, depending on the type of data (large movie files vs a folder full of small images and documents, for example).

 

It's just an in machine test to provide an idea of what the maximum throughput of the disk/unRAID subsystem.

If you get 30MB/s here, you cannot expect anything faster over the network and if a program is reporting higher, then there is something wrong.

 

http://unraid-weebotech.googlecode.com/files/writeread10gb

Share this post


Link to post

Rajahal - thank you tremendously for the great explanation - you have definitely cleared up my misunderstanding.  This is very helpful for an inexperienced unRAID user such as myself.  Perfect for the Wiki (say...the "Newbies guide to how unRAID works"  :)). 

 

With this in mind I now understand where the misconception of having a 7200 rpm parity drive may be of little consequence (unless doing multiple writes as others have mentioned).  With this in mind I may have to rethink moving forward with Caviar Blacks when they come on sale - the power savings with 5400 EADS/EARS are too enticing.

 

Share this post


Link to post

So can we deduce that if we have at least 1 5400 rpm 2TB data drive then a 2TB parity drive that's 7200 rpm would be pointless? I was going to spend an extra £20 for a 7200 on my parity drive but after reading Rajahal explanation I don't think I'll need to now.  :)

 

EDIT: But now after reading this: "For this reason, if possible, use a disk with a higher rotational speed for the parity disk, as it is involved in every "write" to the array.

 

All that said, there are many of unRAID servers with older IDE drives (or "green" drives) that spin at slower speeds, so don't worry if your parity drive spins at 5400 RPM. It will be slightly slower to write to, because it spins slower, but will still be perfectly usable."

 

I'm not so sure anymore!  :P

 

It's only £20, I'll get it anyway.

Share this post


Link to post

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.