Sil 3114 SATA Raid on nforce3 = success from Install

[Snakedriver]

Hi all,
I writing this from a Raid 1 array install on an nforce3-250Gb mobo with Suse 9.1 Pro x86-64 on two mirrored sata drives = success. And, yes, I've rebooted into the array 3-4 times to check. I just followed this guide: http://www.spacebadger.com/main/disp...e=SUSE%20Linux
I printed out a copy & followed it except where it differed from my setup.

From previous experience with suse 9.1 pro, I knew that it contained the sil3114 raid driver, wereas, it does not contain the nforce3 driver -- maybe in a latter kernel upgrade.

The only main thing I've left to do is put a copy of grub on my 2nd array disk to make it a true mirror; but, I'm one happy camper.

Recommend learning about linux partitioning = cylinder count & how much storage it means & such!
I just guessed to set partitions, made both arrays the same though -- but, it worked.

After I screw around with this setup for a while I will probably redo the array to a Raid0 to optimize burst speed cause that's why I got these sata disks.

As you drill down thru the install to partition, click "create custom partition setup" & there is a raid button down at the bottom! I didn't know this til I read the guide. What else don't I know? Anyway, the guide is quite clear about why you need to do what it recommends, when to do it & how.
Enough, Have fun,
Snakedriver
_________________
Desktop = 8rda+/mobile AMD2500+ OC'd
Laptop = HP ze5385us
Suse9.1 Prof on both

[Snakedriver]

don't know what happened to url above, it's:

http://www.spacebadger.com/main/display.php?

itemid=101&topicname=Operating%

20Systems&subname=SUSE%20Linux

[Daniel ~]

Well done Snakedriver! I can fully appreceate your pleasure. In the second week of my Suse 9.1 install!

[Kaitain]

I've had RAID working on both the nforce3 and SIL3114 chips in Gentoo. If you used the SuSE setup utiliity, then you need to be aware that this is not true hardware RAID, controlled by the onboard controller chip, but software RAID.

For this, you only need SATA support, which is present for both the nforce3 and the silicon image chip. Currently there is no reliable hardware RAID support in Linux for either chip.

[Aedan]

Originally Posted by Snakedriver

don't know what happened to url above

Don't worry - the URL is indeed correct. The forum keeps the link correct, but shortens the display so it doesn't take up so much room. Clicking on the link works just fine.

[Snakedriver]

Yes, it's software raid. But, it clearly uses sil3114 driver -- can be seen in yast, hardware, disk controller. Raid1 is a mirror so I don't gain any speed over a stand alone 150 sata drive (vice a 133 ide); that's why I"m gonna redo the array to a Raid0 & hope to gain near a 2x150 speed with parrallel read/write.

Software raid is for we poor folks. Hardware raid cost is beyond my budget.

Aedan, Thanks for the info on the url.

Snakedriver

[Kaitain]

Originally Posted by Snakedriver

Yes, it's software raid. But, it clearly uses sil3114 driver -- can be seen in yast, hardware, disk controller. Raid1 is a mirror so I don't gain any speed over a stand alone 150 sata drive (vice a 133 ide); that's why I"m gonna redo the array to a Raid0 & hope to gain near a 2x150 speed with parrallel read/write.

Sorry, my first post came across rather unfriendly in nature. For the sake of accuracy I wanted to get across the fact that this is RAID controlled not by the SIL chip on board an nf3-250GB based board (I have an EPoX 8KDA3+) but by the Linux kernel. In this regard, you don't even need to buy a RAID-capable motherboard - the kernel will tie together any drive of any type on any bus for the purpose. Pretty useful really

I have 2*Seagate satas (no idea the speed - I'll check dmesg in a few weeks when I reboot) running as a raid0

Do you have hdparm? If not, install it and run the following (as root if you have a high security setting)

 hdparm -Tt /dev/md0

and let us know the speeds you're getting. I'm getting somewhere in the region of 1200MB/s buffered cache reads and 80MB/s buffered disc reads with my Sata-something discs...

[Snakedriver]

Kaitain,

hdparm -Tt /dev/md0 reads:
buffer-cashe 2124 MB in 2.00 Seconds = 1062.16 MB/sec
buffered-disk 6 MB in 0.12 seconds = 51.29 MB/sec

Is this good, bad, fair ???

thanks,
snakedriver

[Kaitain]

Run it at least 6 times. For RAID0 I've had results ranging from

/dev/md0:
 Timing buffer-cache reads:   2064 MB in  2.01 seconds = 1028.04 MB/sec
 Timing buffered disk reads:  246 MB in  3.02 seconds =  81.47 MB/sec

to

/dev/md0:
 Timing buffer-cache reads:   1860 MB in  2.00 seconds = 929.68 MB/sec
 Timing buffered disk reads:  148 MB in  3.02 seconds =  49.03 MB/sec

Software RAID is slightly more state dependent than hardware RAID. It can be faster... a LOT faster. Equally it can run like a dog when the system's busy.

FYI this was all run with Folding@home in the background.

[Snakedriver]

Kaitain,
From raid1:

Originally Posted by Snakedriver

hdparm -Tt /dev/md0 reads:
buffer-cashe 2124 MB in 2.00 Seconds = 1062.16 MB/sec
buffered-disk 6 MB in 0.12 seconds = 51.29 MB/sec

Redid my array to a raid0; reran hdparm 6x & got avg of:
buffer-cashe = 533 MB/sec
buffered-disk = 75 MB/sec.

Gained big time on the disk, but, lost half on cashe?
Is this because I didn't put swap in the array on both disks, i.e., just swap on sba1 outside the array.

I see your's is in thousands; I'm willing to fix this before I finalize this install and press on.

snakedriver

[Kaitain]

Since I have a gigabyte of RAM, I don't have a swap partition at all.

I haven't a clue whether my results here are bad, good or miraculous, since I've not troubled to find out how it compares with other users' RAID arrays.

The cache read doesn't actually read from the disc at all, but instead from the cpu/memory subsystem. If you're getting half the speed consistently on cache reads, then the issue lies not with your discs, but with your RAM. Compare your specs to the one in my signature (system 3) and see whether you can spot any obvious differences. Software effects will play a part, too. My system is running Gentoo with a fairly well optimized kernel (gotta love tweaking cflags ) whereas SuSE's default offering is designed to be as broadly compatible as possible, possibly at the expense of speed.

From the info pages of hdparm:

-T Perform timings of cache reads for benchmark and comparison pur-
poses. For meaningful results, this operation should be
repeated 2-3 times on an otherwise inactive system (no other
active processes) with at least a couple of megabytes of free
memory. This displays the speed of reading directly from the
Linux buffer cache without disk access. This measurement is
essentially an indication of the throughput of the processor,
cache, and memory of the system under test. If the -t flag is
also specified, then a correction factor based on the outcome of
-T will be incorporated into the result reported for the -t
operation.

-t Perform timings of device reads for benchmark and comparison
purposes. For meaningful results, this operation should be
repeated 2-3 times on an otherwise inactive system (no other
active processes) with at least a couple of megabytes of free
memory. This displays the speed of reading through the buffer
cache to the disk without any prior caching of data. This mea-
surement is an indication of how fast the drive can sustain
sequential data reads under Linux, without any filesystem over-
head. To ensure accurate measurements, the buffer cache is
flushed during the processing of -t using the BLKFLSBUF ioctl.
If the -T flag is also specified, then a correction factor based
on the outcome of -T will be incorporated into the result
reported for the -t operation.

For comparison, the PATA drive in my system is a slow 5400rpm Seagate U6 series ATA100 drive:

/dev/hda:
 Timing buffer-cache reads:   2292 MB in  2.00 seconds = 1145.60 MB/sec
 Timing buffered disk reads:   80 MB in  3.00 seconds =  26.63 MB/sec