Poor Mans RAID and Disk Tips

[WarDad]

I started this piece of work to satisfy my own curiosity. Last year I wanted to RAID 0 on my new AMD 965 system, but ended up dual booting because W7-64 was too young. It's going to be a Christmas gift this year. I already have this years Sandy Bridge system up and running.

What a mix:
Crucial M4 SSD,
Two 1TB WD Caviar Blacks,
Two 2TB Hitachi Cool Spin 5K3000 green drives,
AMD SB850 + GigaByte SATA controllers, on a GA-890GPA-UD3H
Intel South Bridge and Marvell SATA controllers, on a ASUS P8P67 Deluxe
One Lab Rat professional IC evaluator

I present to you a link to "Poor Mans Raid and Disk Tips.pdf"

http://www.mediafire.com/?2gcb4531pz5x8yd

Comments anyone?

 

[WyomingKnott]

One comment: post a link to the pdf, not Mediafire! I get more than enough ads in my life.

 

[4745454b]

I don't like either. I do my best to avoid .pdf as a whole. To many exploits. Some newcomer on a forum wants me to read a .pdf? PASS.

 

[WarDad]

This forum did not have an obvious file attachment option.
Mediafire file hosting was free. Not pleasant, but free.

OK, the pdf is 11 pages long and has inline images side by side.
I'm not sure if the moderators want it all posted, or exactly how to do it.

I like pdf documents. I like having reference documents at home.
I hate detailed online game strategy and info that is web only.
Give me both. But, then I'm older and can live without the web.

 

[JackNaylorPE]

Since Im not running a huge database with multiple user access, or tying to speed storage subsystem up to raise a benchmark, it's hard to find a reason to run RAID 0

 

[WarDad]

Chapter 1 - Better Disk management, Intro to benchmarks, verifying AHCi-NCQ and Write Cache

 

[WarDad]

 

[WarDad]

 

[WarDad]

Chapter 2 - Why RAID when you can SSD?

 

[palladin9479]

Two 1TB WD Caviar Blacks,
Two 2TB Hitachi Cool Spin 5K3000 green drives,

I can understand using the SSD for OS but why two separate RAID's on a single system?

If it's storage you can do 4x1GB or 4x1.5GB in RAID 0 / 10 / 5 depending on how you feel.

 

[WarDad]

I had two sets of drives to play with and benchmark. I never had both setup side by side.

 

[WarDad]

Chapter 3 - RAID 0, RAID 1, and other Motherboard RAIDS

 

[WarDad]

 

[WarDad]

End Chapter 3

 

[palladin9479]

I had two sets of drives to play with and benchmark. I never had both setup side by side.

Ahh ok makes sense.

 

[WarDad]

Chapter 4 - Intel P67 mixing one SSD and HDD RAID.

 

[WarDad]

Chapter 5 - Changing SATA drivers on XP boot drive.

 

[4745454b]

To be completely honest I'm not sure you can draw any correct conclusions with this data. You are using very different drives in your testing. If you were testing the same drives on different controllers then you would have accurate data from which to draw conclusions.

 

[WarDad]

The Black and Green HDD were close enough in performance to allow a sparse test matrix. A strict test protocol with a full test matrix would have been too large.
The trends and tips are the point of the ariticle. Ex. The reader is better served when they can tell difference between 4K and 4K QD32, and update their drivers to AHCI-NCQ.

 

[4745454b]

Rotation speed is the same? Platter density? Cache? Controller microcode? I get the feeling they are a bit more different then you think. And I'm not sure you or I have enough knowledge about them to know if the difference in scores your seeing is the test or the drives.

 

[WarDad]

I can tell the difference, that's why I'm paid the big bucks. This isn't about the best benchmarks. I used brief and simple benchmarks that almost anyone could use. The Green drives get a 4K read around 0.6MB/s, and the Blacks get around 0.7MB/s. Both suffer a little if they are boot drives. The Blacks have less seek time, more platters (cylinders), and higher rotational speed. The Green ones are a year newer, have higher disk density, and Hitachi is known for good Sequential read speeds.

Read Wiki NCQ to get a better understanding of why 4K and 4K QD32 read results differ.
Obviously if 4K QD32 read is not significantly better than 4K, the reader is not getting full performance.

The Gigabyte controller also supports legacy Floppy and PATA (ribbon cable) drives.
It's driver has a check box for legacy Tagged Command Queueing.
NCQ (Native Command Queueing) does not appear to work here. Coincidence? No.

In the optimize chapter I did a test case with Write Cache On vs. Off. With off showing a drop in 512K and Sequential tests write speeds.
The GigaByte and Marvell secondary SATA controllers also had write speeds drop off on 512K and Sequential tests when pushed by RAID.
Both secondary controllers PCIe 2.0 x1 serial channels rated about 500MB/s.
Both SouthBridge controllers have PCIe 2.0 x4 serial channels for 4x the bandwidth. They did not have a write slowdown.
Why? Who cares, I can benchmark it and avoid using the secondary SATA controllers.

 

[4745454b]

And I do like how you fail to mention that depending on your board those "secondary SATA controllers" might be the only 6Gbps ports on your board. Your SSD should be plugged into them. But of course you said to never use them, so SSD speeds will suffer.

 

[WarDad]

How does "I can benchmark it and avoid using the secondary SATA controllers." become "never use them"? I think you missed the "I" and the "benchmark" parts.

BTW: SSD speeds do suffer on my Marvell secondary controller. Maybe a PCIe 2.0 x4 Sata controller card (bootable) would be a better choice.

 

[4745454b]

I think I wrote what I said because of your last paragraph in that first picture post, though its not as hard a "never" as I remember reading. I've looked back through this but can't seem to find where I saw what I remember seeing. I'm also not sure the issue which you speak of has anything to do with "pci bandwidth". 6Gbps ends up topping out around 600MB/s. 4x lanes of PCIe bandwidth ~1GB/s, so the bandwidth is there. Assuming of course that secondary chip is hooked up to 4x lanes. I haven't bothered to check, only going from what you wrote.

 

[WarDad]

Chapter 6 - The Test systems and benchmarks

Hard Drive Size Matters
A 2TB total capacity is compatible with standard MBR boot drive formatting. The new GPT formatting supports larger data drive capacities but it's not always backwards compatible. If your RAID exceeds 2TB then do some homework at your motherboard and OS forums. Your SATA controller may not support GPT, but it might provide a work around by partitioning multiple RAID 0 drives on the hard drives.


The Test Hardware
Drives:
2 each 1TB Western Digital Caviar Blacks, 3GB SATA, 7200 RPM, WD1001FAES
2 each 2TB Hitachi Cool Spin Green drives. 6GB SATA, ~5900 RPM, HDS5C3020ALA632
1 each 256GB Crucial M4 SSD, 6GB SATA, CT256M4SSD2
Last years system:
Gigabyte GA-890GPA-UD3H rev. 0, XP OS, AMD 965 BE, RADEON HD5870
AMD 850 South Bridge SATA Controller is very flexible.
Gigabtye controller supports SATA, legacy PATA, and Floppy drives.
This years system:
ASUS P8P67 DELUXE, W7-64 OS, I7 2600K 3.4 GHz, 2x GTX560TI, Crucial M4 SSD
Intel South Bridge SATA controller,s bios is all RAID or none.
Marvell SATA controller does not support boot drives.
Backup system:
VANTEC NST-300SU-BK Hard Drive enclosure with a 2TB green drive. It's an easy to use removable storage connected by USB 2.0 or eSATA.


The Bench Marks

IOmeter is an old program is for benchmarking servers. Intel gave it away many many moons ago. I first used it to test two independent disk simultaneously on one or two SATA controllers. It is flexible, has quite a learning curve, and produces spread sheet numbers. I became proficient and the results were close to Crystal Disk Mark 3.

Crystal Disk Mark 3 is a free program that is informative, works well, is easy to use, and is pleasing to the eye. Other programs may do more or work better but to a poor man free is very, very good.

HD Tach is free and gives a unique picture of performance across the hard drive platter. It also measures seek times. The Hitachi green drives were 19.3 ms and the WD Blacks were 13.6 ms.

I noticed the read and write times were a little off for OS boot drives. Possibly the Bench Mark program and OS are competing for disk access. The OS also occupies the fastest real estate on the hard drive.