0

I recently upgraded(because of fear of failure) my back up drive from a 750GB Toshiba Canvio USB3 drive(couple of years old) to a WD My Book Duo 6TB USB3 drive. I configured WD My Book Duo as Raid 1 for redundancy. These are the benchmark numbers I got using CrystalDiskMark.

WD My Book Duo 2x3TB in Raid 1 configuration

-----------------------------------------------------------------------
CrystalDiskMark 6.0.0 x64 (C) 2007-2017 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :   154.008 MB/s
  Sequential Write (Q= 32,T= 1) :   154.307 MB/s
  Random Read 4KiB (Q=  8,T= 8) :     0.618 MB/s [    150.9 IOPS]
 Random Write 4KiB (Q=  8,T= 8) :     1.448 MB/s [    353.5 IOPS]
  Random Read 4KiB (Q= 32,T= 1) :     0.595 MB/s [    145.3 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :     1.512 MB/s [    369.1 IOPS]
  Random Read 4KiB (Q=  1,T= 1) :     0.562 MB/s [    137.2 IOPS]
 Random Write 4KiB (Q=  1,T= 1) :     1.435 MB/s [    350.3 IOPS]

  Test : 1024 MiB [D: 1.0% (26.6/2794.4 GiB)] (x5)  [Interval=5 sec]
  Date : 2017/11/07 10:16:56
    OS : Windows 7 Enterprise SP1 [6.1 Build 7601] (x64)

Toshiba Canvio 750GB

-----------------------------------------------------------------------
CrystalDiskMark 6.0.0 x64 (C) 2007-2017 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :    12.058 MB/s
  Sequential Write (Q= 32,T= 1) :    96.205 MB/s
  Random Read 4KiB (Q=  8,T= 8) :     0.403 MB/s [     98.4 IOPS]
 Random Write 4KiB (Q=  8,T= 8) :     1.108 MB/s [    270.5 IOPS]
  Random Read 4KiB (Q= 32,T= 1) :     0.320 MB/s [     78.1 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :     1.055 MB/s [    257.6 IOPS]
  Random Read 4KiB (Q=  1,T= 1) :     0.263 MB/s [     64.2 IOPS]
 Random Write 4KiB (Q=  1,T= 1) :     1.126 MB/s [    274.9 IOPS]

  Test : 1024 MiB [E: 25.7% (179.8/698.6 GiB)] (x5)  [Interval=5 sec]
  Date : 2017/11/07 11:10:22
    OS : Windows 7 Enterprise SP1 [6.1 Build 7601] (x64)
  1. In sequential RW performance WD (even in Raid 1) is way better compared to Toshiba. However when I tried copying around 4GB of photos ~4000 files Toshiba was more than twice faster(~2minutes completion) compared to WD(~5minutes completion). Is this normal? Is the Raid 1 overhead of WD cause of this performance drop?
  2. How do I read data from the hard drive without using WD controller(The box)? i.e by plugging the drive directly into the SATA port of my computer. Is there WD software which can help me recover it?
  3. Should I go for a backup of backup instead of using Raid 1 backup since real world performance of my Raid 1 setup seems to be low?
  • I don't understand the question. Are you asking about RAID performance or data backups? – joeqwerty Nov 7 '17 at 17:13
  • Both. i.e is performance hit part of raid 1? Should I consider raid 1 backup instead of a manual backup of backup? – Jean Nov 7 '17 at 17:20
  • A good backup is two independent copies on two different media. Using a RAID 1 for the two copies makes them dependent, the RAID controller makes it a single medium. – Zac67 Nov 7 '17 at 18:31
1

RAID 1 shouldn't have any noticeable performance overhead. Each write results in two physical write operations, but the hardware should be doing these in parallel.

Benchmark programs typically use a large number of I/O threads, each running multiple I/O operations, in order to keep the disks as busy as possible. The goal is to ensure that the drives' command queues are full ("Q=32" in your logs) and that they are pushed as close to their max as possible. Benchmarks frequently use low-level I/O operations that bypass things like the OS's filesystem layer.

File copy operations - especially those done through a typical file manager - tend to be single-threaded processes that only have one operation queued at a time. They're optimized for reliability, not speed, and have enough extra overhead that they don't come close to demonstrating the drive's true performance. File copy performance can also vary wildly when running the same test multiple times. I wouldn't read too much into the performance difference in your file copy tests. You may get better performance using a dedicated backup utility that can write multiple files at once and keep the drives busier.

In general, expect that a RAID volume can only be read on a controller similar to the one that configured it. I've never encountered a hardware RAID controller that had the capability to access its volumes via software over a normal HBA (even though this would be incredibly useful). If your enclosure/controller dies, I would expect that you could still transfer the drives to another WD box to regain access.

Whether to use a backup-of-backup configuration or a RAID 1 will depend on what you are trying to protect against. If you're most concerned about a drive hardware failure, RAID 1 might be the better choice. If you're more concerned about file corruption, accidental deletion, or virus damage, a backup-of-backup configuration might be better (it can take longer for changes to propagate to both backups, giving you more time to detect the problem and fix it).

| improve this answer | |
6

RAID - no matter what level - is not a backup. It is hardware redundancy and is meant to mitigate any loss of either data OR run time/uptime due to sudden catastrophic hardware failure.

RAID will not protect against accidentally deleting the wrong file, redirection of the wrong output in the wrong way to the wrong file, uninstalling the wrong package, overwriting the contents of one file with another, the effect of a virus or trojan (file removal, corruption, or ransomware encryption), etc. That is what backups are for.

Now, that said, I'll admit to not using backups, or if I do it is a one-off backup of "gee, its been a year since I've saved my mozilla bookmarks, better export that out and upload it to my VPS", or "I just finished this final project for a programming class, I'll tar it up and put a few copies on other machines". BUT I also run RAID-1 for my /home partition on both my work and home desktop machines. Why? Simple - I've lost more data over the years to "just lost power and my hard drive is now dead" than I've lost data to "where did I leave that floppy" or "I got a virus" or "I deleted the wrong file"

| improve this answer | |
  • I didn't say that RAID is a backup. My question was whether using two different backup drives or one raid 1 backup setup was better. – Jean Nov 7 '17 at 20:33
2

You're confusing two completely different approaches of storing data. Firstly, and I cannot stress this enough, RAID is not a backup.

RAID is a mechanism for guarding against disk failures by storing data on two or more disk drives. Redundant Array of Independent Disks.

Backups are duplicating your data from its source, in order to keep a copy for disaster recovery purposes. To be true backups, they need to be wholly independent of the source media - how independent is your choice, but at a minimum a good backup should be on different media/storage, ideally in a different location and most ideally in a different continent.

RAID guards against disk failure, backups guard against a whole host of different failures - for example natural disaster (fires, floods), human error (whoops, I deleted my master budget file) or theft.

Now, with all of this in mind, performance should not be your primary concern, and you should instead be considering your primary purpose of your storage, is it for redundancy or for backups?

| improve this answer | |
  • No. I am not. I am talking about adding redundancy to my backup drive. – Jean Nov 7 '17 at 20:29
  • Both approaches are valid, as the purpose of a backup is not altered by making the underlying storage redundant or making multiple copies (although a backup of a backup is not a backup unless it is stored in a different geographic location). – Craig Watson Nov 7 '17 at 22:26

Not the answer you're looking for? Browse other questions tagged or ask your own question.