Is Software Raid 5 right for me?

[Codesmith]

I have 4 empty WD 120 GB hard drives in good health (DST, SMART) but all over 3 years old and it's a safe bet that not all four are making it to 5 years.

One is gone into an external USB/SATA enclosure, so that leaves three. One too many for mirroring, one short for 1+0, so I thinking Raid 5.

Today I hacked XP and put the three drives into a RAID 5 volume.

(Before anyone says anything, I can't afford a good RAID 5 card and the cheap ones RAID 5 performance is generally much worse than Windows Raid 5)

Now that I have setup software RAID 5 I am wondering whether I just increased or decreased the safety of my data!

Anyone out there have any real world experience good or bad with Windows Software Raid 5?

Any potential problems that would make it less reliable than simply keeping my data on three seperate aging hard drives?

 

[Codesmith]

The thing is my Antec 550 watt is dying (its 5v kept dropping and is currently being outdone by the Enlight 340 watt I swapped in from my 2nd PC!), and I just bought a new digital camera & X-Fi ExtreemMusic.

So no $$ for computer parts just now, and when I have it its going toward a good powersupply not a RAID card .

So far I am going to do the following to test the reliablity.

1) Restart with one drive uplugged & test the rebuild.
2) Restart with two drives unplugged and see if everything is ok once I reboot and connect the other two drives.
3) Restore from a backup without the Raid 5 hack and see if being in an unhacked system does permanent damage.
4) Move the drives to different controllers, and then to a different PC.

I am having trouble finding a program that will do a transfer rate test on a Software Raid Array. I think I am going to rip a DVD to my HD, defrag mount it with deamon tools and use Nero's CD/DVD Speed.

That should at least give me a % difference between a single drive and the RAID5 array.

I am guessing that the RAID5 is slower, but thats ok if it means that when one of these drives dies I just buy a replacement with 0 data lost.

 

[Codesmith]

I did a lot of testing.

Very Reliable (could not manage to lose data despite swithing controllers, disconnecting drives at start up ect).

Very Slow Read performance.

Extreemly slow write performance.

So slow in fact that I decided to switch to Nvidia Raid 1.

 

[Codesmith]

CPU utilization wasn't terribly high, now I am curious about where the bottleneck lies.

Maybe the data just needs to get pushed arround and crunched too much before it makes it to/from the hard drive.

Maybe there is some sort of artificial limit placed on how much CPU time software RAID 5 can grab?

 

[Codesmith]

HD Tach works at a lower level than XP Software Raid 5 and sees and benchmarks the array as individual drive. Maybe thats a clue as to why XP's Raid 5 is so slow. Then again all cheap Raid 5 seems to be very slow.

Here are my benchmarks.

I have 5 drives in that PC, A SATA 74 GB Raptor, 3 PATA 120 GB WD-1200BB's (2mb cache) and 1 PATA WD-1200JB (8mb cache). I have an NF4 Ultra MB (MSI K8N Neo4 Platinum) and an Promise Ultra ATA 100 card which gives me enough IDE channels to avoid having slaves when testing XP's RAID 5

The Raptor has a 30 GB partion hosting my OS. This is the fastest part of the hard drive so the benchmarks so it benchmarks a bit better than if I had one 74 GB partion.

Not shown are the results that show the matched 120's vs the mismatched 120's in Nvidia Raid 1 configuration.

Also the NVIDIA 1+0 results have to be taken with a grain o f salt as NF4 only offers 2 PATA channels so half the array was slave to the other half! The results might be quite different if two of the drives where SATA.

Also note that NVIDIA Raid 0 increased HD Tach's results but slightly lowered the scores on PCMark05.



PC Mark 05 1.0.1 74 GB Raptor C:
HD Tach
XP Startup 10.385
Application Loading 9.933
General Usage 8.489
Virus Scan 76.825
File Write 66.712

Single WD120
HD Tach
XP Startup 7.52
Application Loading 6.606
General Usage 5.559
Virus Scan 45.873
File Write 41.467

Nvidia Raid 0 2HD
HD Tach 67.7
XP Startup 5.959
Application Loading 5.805
General Usage 4.911
Virus Scan 45.611
File Write 29.862

Nvidia Raid1 2 HD
HD Tach 19.9
XP Startup 5.724
Application Loading 4.595
General Usage 3.77
Virus Scan 23.026
File Write 15.483

Nvdia 1+0 w Slaves
HD Tach 23
XP Startup 4.895
Application Loading 4.097
General Usage 3.624
Virus Scan 21.734
File Write 15.048

Nvdia 1+0 w Slaves
HD Tach 23
XP Startup 4.895
Application Loading 4.097
General Usage 3.624
Virus Scan 21.734
File Write 15.048

XP 5 3HD
HD Tach
XP Startup 1.46
Application Loading 1.224
General Usage 0.827
Virus Scan 19.397
File Write crash

XP 5 4HD All Master
HD Tach
XP Startup 1.445
Application Loading 1.252
General Usage 0.844
Virus Scan 19.691
File Write crash

XP 5 4HD w Slaves
HD Tach
XP Startup 1.277
Application Loading 1.144
General Usage 0.787
Virus Scan 12.813
File Write crash

 

[Codesmith]

Thats all I have.

HD Tach DOES NOT work with XP's Software RAID and I don't know of a utility reporting those values that does.

HD Tach just see's individual drives.

PCMark05 and WinBench 99 are the only alternatives that see disks as a Raid 5 array, but they don't measure CPU utilization.

So for CPU utilization all I got had was task manager and how the system feels .

The system didn't feel at all sluggish during drive regeneration or copying folders and I was able to game just fine. But I didn't delibrately set out to judge CPU utilization

With Nvidia Raid 1 I can't game durring long transfers and I bounce between 100 and 80% cpu utilization.