Hard Drive Configuration Comparisons Stock VS WD Black VS RAID

[minifridge1138]

So for black Friday I got a WD 1TB Black drive.
I have the original 640 GB Hitachi drive that came with my mac, a 1 TB Hitachi drive that I originally bought. Rather than add more drive space, I thought I would setup a RAID1.

I tried three different setups and here are the results:
-Original Apple/Hitachi 640 GB
--Disk Test -------------------- 64.21
---Sequential ----------------- 133.76
----Uncached Write ----------- 144.10 88.48 MB/Sec [4K blocks]
----Uncached Write ----------- 149.10 84.40 MB/Sec [256K blocks]
----Uncached Read ----------- 97.76 28.21 MB/Sec [4K blocks]
----Uncached Read ----------- 165.76 83.31 MB/Sec [256K blocks]
---Random--------------------- 42.25
----Uncached Write ----------- 15.43 1.63 MB/Sec [4K blocks]
----Uncached Write ----------- 116.73 37.37 MB/Sec [256K blocks]
----Uncached Read ------------ 70.67 0.50 MB/Sec [4K blocks]
----Uncached Read ------------139.86 25.95 MB/Sec [256K blocks]


-WD Black 1 TB
--Disk Test --------------------- 94.83
---Sequential ------------------ 147.17
----Uncached Write ------------ 157.65 96.82 MB/Sec [4K blocks]
----Uncached Write ------------ 157.84 89.31 MB/Sec [256K blocks]
----Uncached Read ------------ 102.76 30.07 MB/Sec [4K blocks]
----Uncached Read ------------ 209.33 105.21 MB/Sec [256K blocks]
---Random --------------------- 69.95
----Uncached Write ------------ 25.35 2.68 MB/Sec [4K blocks]
----Uncached Write ------------ 310.33 99.35 MB/Sec [256K blocks]
----Uncached Read ------------ 109.54 0.78 MB/Sec [4K blocks]
----Uncached Read ------------ 185.68 34.45 MB/Sec [256K blocks]


-RAID 1 across WD & Hitachi 1 TB
--Disk Test --------------------- 56.93
---Sequential ------------------ 129.38
----Uncached Write ------------ 170.13 104.46 MB/Sec [4K blocks]
----Uncached Write ------------ 149.71 84.71 MB/Sec [256K blocks]
----Uncached Read ------------ 74.28 30.07 MB/Sec [4K blocks]
----Uncached Read ------------ 204.33 102.62 MB/Sec [256K blocks]
---Random --------------------- 36.49
----Uncached Write ------------ 11.25 1.19 MB/Sec [4K blocks]
----Uncached Write ------------ 142.01 45.46 MB/Sec [256K blocks]
----Uncached Read ------------ 115.43 0.82 MB/Sec [4K blocks]
----Uncached Read ------------ 199.79 37.07 MB/Sec [256K blocks]


Conclusions:
-The WD Black beat the Apple/Hitachi in every single category.
-The RAID was marginally faster than the WD Black at some tests, and significantly slower at others (particularly the uncached write of 256 blocks).
-The RAID was faster than the the Apple/Hitachi at every test, except uncached random writes in 4k blocks.

My personal thoughts and why I'm still running the RAID setup because:
1) It provides protection against hardware failure
2) It was faster than the 640 I was using until a week ago.
3) It out performed the Black in some read tests, and was usually very close on others (except where noted above).

Enjoy!!!

 

[Honumaui]

what are you backing up to ?

raid is never a backup but also I would never run raid 1 with two drives that are so different ?

why not just use the new as storage and bu to the old one ?

 

[diazj3]

Nice analysis - I was just wondering this myself, since I set raid yesterday on my MP.

What did you use to measure the speeds?

 

[Honumaui]

looks like Xbench ? but I would go get AJA and test with that

 

[nanofrog]

raid is never a backup but also I would never run raid 1 with two drives that are so different ?

why not just use the new as storage and bu to the old one ?

Assuming there is no backup, then this would be a better method IMO as well, as it avoids the immediate duplication of problems as happens in RAID 1 (i.e. user error such as accidentally erasing a file).

If it's used as an OS/applications volume only, a backup may not be absolutely necessary, but a separate disk (clone actually), is a really nice thing to have (less effort, and usually time to do a restoration - change it to the boot location, and done). From what we've covered before, I've the sense you'd agree with this.

As per disks, though it's nice to have identical disks (less capacity loss, and predictable performance), it's not a necessity.

looks like Xbench ? but I would go get AJA and test with that

Running AJA would be a good idea.

 

[minifridge1138]

Yes, they results are from XBench.

I will look into AJA, but I can't do a test of the WD Black as a sole boot because it is now part of my RAID array.

As for why i did it:
1) Challenge: to see if I could
2) Curiosity: to see what difference it would make

I agree that using identical drives would be better (preferably from different manufacturing batches); however, the drives' key stats are the same: 1 TB capacity, 7200 RPM, 32 MB cache. Architecture and firmware are different, and I understand that makes a huge difference, but it isn't like I setup a Raid between a WD 10,000 rpm Raptor and a 5400 RPM external USB drive (saddly i have seen that done).

Right now, i still have the 640 GB drive with the OS installed on it and can use that as an emergency backup in case of catastrophic loss.

I'm really going to split hairs on this, but technically a raid1 is backup.

A RAID1 is a real-time backup (data duplication) of all data written to that volume. Because data is being written to 2 disks, the writes are slower (in theory by 1/n where n=number of disks in the array) but reads should be faster (in theory by a factor of n) because rotational latency has been reduced by the introduction of additional disk heads, additional cache, etc.

Time machine is more of a version control system (similar to svn, cvs, git, etc) than a pure backup (the two are not mutually exclusive). It is setup to automatically commit at regular intervals. So it is a regular backup AND a history of all changes made.

The biggest difference is that if Time Machine Runs every hour, and your machine dies at X:45, then you can restore to X:00 and you've lost 45 minutes worth of file changes.

If one disk in your RAID dies at X:45, the other drive is fine and you just keep going. But since it is only the latest version, you are not protected against data corruption, accidental deletion, etc.

I say that Time machine is more Version Control than backup, because a traditional backup systems take a complete, independant backup of the system every time. So if you are using 600 GB of disk space, every backup is 600 GB in size. The backups are to separate volumes (tape backups are still very common) and are completely agnostic of each other.

Time Machine's first backup is a complete backup of the entire disk (so 600 GB in this example). The second time it runs (1 hour later) it does not store 600 GB of data; rather, it does a diff between the disk and the backup and only stores the changes (so maybe only a few MB). The next time it runs, it does a diff between the disk and the latest stored copy (so maybe only another few MB). This allows you to conserve disk space, but means that in order to restore you must have an image of the data to which you can apply the diffs. The diffs by themselves are useless.

Questions, comments, crude remarks?

 

[nanofrog]

I'm really going to split hairs on this, but technically a raid1 is backup.

No it's not.

RAID 1 was designed for high availability (24/7 operation without having to shut the system down and replace the volume, such as OS/applications use in servers). Such systems must to be able to tolerate a failure without operational loss. It functions by automatically duplicating data from member 0 to member 1, so any corruption, user error, .... will be duplicated (besides user error, data corruption can still cause a RAID 1 to have to be rebuilt from backups <think a file being written on bad sectors, and the drive is no longer able to be re-mapped>).

A backup is an independent/separate volume from the primary data location (no "connection" between it and the primary volume it's receiving data from). So you still have the data, even if the RAID 1 (or any other disk configuration from single to the most redundant storage design possible) fails. And this can, and does happen.

Hence the statements that you still need a proper backup. You can use a RAID to backup a RAID, but again, the "connection" is only through backup software, not the RAID function itself (but this is still a backup, not automatic duplication performed by the RAID function itself).

 

[Honumaui]

pretty much Ditto Nanofrog and no crude remarks heheheh


but in my words

Raid 1 is not a BU
in my book and some buddies from Colorado Memory Systems I was friends with no raid is a BU !!! you can use Raid for BU but raid itself is not


yes you can use different drives I was just stating I would not ? JBOD I would but not raid 1 ? might as well with how cheap drives are use the same




TM is incremental BU
CVS is something different ?
wiki both to get better answers but their is a difference in the two as I have used a CVS system heavily in the past and its not backup