The 03.07.17 at 00:47, Philipp Thomas wrote:

There's *nothing* in these cheap RAID controllers (sometimes called fakeRAID) that you can't achieve with the Linux software RAID driver, unless you want to run M$ on the same system. These controllers are not much more then simple ATA controllers with additional functionality in the BIOS. All of the RAID logic is implemented in the driver, i.e. in software.

This interests me. Do you know of "further reading" material about it? Howtos, webpages? -- Cheers, Carlos Robinson

On 2003.07.18 20:40 Philipp Thomas wrote:

...

This interests me. Do you know of "further reading" material about it? Howtos, webpages?

I'm sorry, but I don't have anything at hand. It's just what I gathered over the years from making support and following discussions both internal to SuSE and on lists like lkml.

Ah! I was afraid that would be the case: each of us has his own area of expertise which is difficult to find documented. They call it experience, know-how... and this is a fast changing field.

BTW, the pdcraid and hptraid drivers in the kernel aren't driver in the normal sense but nothing more than md personalities, i.e. they make the disk layout the BIOS created understandable for the md (software RAID) driver. The real work is done by the software RAID driver.

Ok, then lets try to synthesize the ideas I'm getting. 1) Cheap raid cards, or motherboard things, are not really hardware raid solutions. 2) They need support from the OS as a driver, be it windows, Linux, whatever. 2b) They don't work as raid without such drivers 4) A "pseudo hardware raid" like this, with the proper personality, can (or could) be used both from windows and Linux - of course, if formated with a filesystem both understand. 5) They are probably less flexible than a Linux software raid. 6) They are perhaps faster :-? Am I approximately correct? Incorrect? -- Cheers, Carlos Robinson

On Sun July 20 2003 5:39 pm, Philipp Thomas wrote:

"Carlos E. R." [20 Jul 2003 15:32:44 +0200]:

...

1) Cheap raid cards, or motherboard things, are not really hardware raid solutions. 2) They need support from the OS as a driver, be it windows, Linux, whatever. 2b) They don't work as raid without such drivers 4) A "pseudo hardware raid" like this, with the proper personality, can (or could) be used both from windows and Linux - of course, if formated with a filesystem both understand. 5) They are probably less flexible than a Linux software raid. 6) They are perhaps faster :-?

Am I approximately correct? Incorrect?

Yes, this is mostly correct.

ob 4) Both OS's will see the actual RAID array just like normal drives/partitiions. But each OS can only use that part of the RAID array that's formatted with a filesystem it supports.

ob 6) I can't tell. I've never dealt with those things (at least not as RAID array). If I needed RAID, I'd use a 3ware card. That's real RAID in Hardware at affordable prices.

Philipp

From personal experience, but without using a stopwatch and comparing two separate installations, RAID (with apologies to Philipp for abbreviating this from "Not Really Real RAID"), I would only say it seems not to be slower. It is clearly not the difference in getting a new MB/CPU and cetera. Test results I had seen earlier (Anandtech, maybe?) suggested a performance increase of maybe 10% if you were lucky in striped mode, maybe a couple of % if anything in mirrored mode, though reads were a little quicker than writes. Personally, I do not see (Philipp's hated) RAID as a solution to performance issues, but rather as a solution to other issues, eg, lots of old disks hanging around, each too small to use alone, and so on. Regards, Richard

The 03.07.20 at 18:05, richard wrote:

From personal experience, but without using a stopwatch and comparing two separate installations, RAID (with apologies to Philipp for abbreviating this from "Not Really Real RAID"), I would only say it seems not to be slower. It is clearly not the difference in getting a new MB/CPU and cetera. Test results I had seen earlier (Anandtech, maybe?) suggested a performance increase of maybe 10% if you were lucky in striped mode, maybe a couple of % if anything in mirrored mode, though reads were a little quicker than writes. Personally, I do not see (Philipp's hated) RAID as a solution to performance issues, but rather as a solution to other issues, eg, lots of old disks hanging around, each too small to use alone, and so on.

Well, supossedly, raid5 is faster. I have a software raid1 (mirror) partition here, just for testing (I don't have any type of hardware or pseudo hardware raid), and it seems way slower than the rest of the non raid partitions. Let me see... (time dd_rescue /backup/boot2003.gho /dev/null) The raid partition reads at 20508kB/s Average (big file). Another partition reads at 24014kB/s. A write of one 204 Mb file times as: 7804kB/s (raid) 15995kB/s (normal) For these tests, not very scientific really, I used "time dd_rescue Mail.tar.gz ./delete" - because dd_rescue outputs some statitistics, although it is slow transferring. But it seems that write operations on a sofware raid 1 (mirror), two disks array, is half as slow in writing, and 25% slower when reading. The reading part I thought would be faster at least... In my case, both disks share the same cable (the other one is used for the cdwriter and the dvd - the CD forces the HDs to be terribly slow if placed on the same cable: no UDMA5 then). Probably the results would be different if I had the HD on separate cables, and perhaps even better as four disks as raid 5 on four different cables - but I don't have the hardware to test that, and I'm not buying it just for testing ;-) -- Cheers, Carlos E. R.

On Mon July 28 2003 2:22 pm, H du Plooy wrote:

On Wed, 2003-07-16 at 22:26, Richard wrote:

...

Does anyone have any opinions (really opening up here) between the MSI K7N2Delta ILSR motherboard and the Gigabyte 7N400 Pro? Both seem to be good

Let us know what you got and how it works. I'm interested in ditching my old SD-RAM board for a DDR setup so my Athlon can perform the way it should. I've seen the benchmarks and the nForce boards rock, but I'm sceptical of nVidia under linux (after two years of frequent unneccessary trouble with various nVidia cards.

Thanks Hans

I got the Gigabyte board and my review would be mixed at this point. The CD was supposed to have linux drivers on it and it did not. I downloaded the linux files from the Gigabyte website, but they only supplied files for RedHat and Mandrake. Thus the included drivers, as modules, were built for those two distros. Simply copying them did not work as my system did not accept them since it recognized the RH/Md indentifying info. I downloaded a driver file from the Nvidia website, after some searching and they seem to have worked ok for the most part. Sound seems to work for system sounds but I had to do some "monkeying" to get KSCD to play. I understand that that is not unusual so I cannot attribute that to the G-MB/NV drivers, yet. The ethernet driver was another story as I had to do some fiddling to get it to work with Yast. Yast recognized the eth controller but needed to be told to use the "nvnet" module and it had to be insmod'd and added to etc/modules.conf. I wanted the availability of the Raid chip, more for the ability to use more than four IDE drives than necessarily for the raid capability. Again, drivers are on the website for RH and Md only. I was able to download the driver source files from the ITE website (makers of the Gigaraid chip), after registering on that site, and built (I think) a driver module that my system accepted. I have not tried to use it in a raid setup yet but my system does recognize and can mount the drives I have attached to that controller, after copying the created module to the /lib/modules/... and insmod'ing. The installation instructions for a raid setup direct to replace the installation kernel with their supplied kernel, which kernel is not available for SuSE, of course. My concern is that the installation instructions mention that the supplied (supposedly) kernel already has support for the ITE chip built in. Since I do not see support for that chip in the standard kernel I am too queasy and too much of a 'fraidy-cat to go much further. Telephone call to Gigabyte's tech support said they will not provide support for SuSE for the raid chip because they do not have the time to try out the chip with all distros, and cetera, and the usual.... They suggested I try the ITE people. I did and their tech support said similar, they only had time and resources to do the two distros they did. Disappointing. On the positive side, the board is up and running and seems nice. On the tremendously negative side I have not been able to use the dual-channel DDR memory capability. I have two memory sticks, which test out fine alone and together, so long as they are in same channel. When placed into different channels (a requirement for dual-channel) they system shows multiple errors and locks up, though at different places for both WinXP and SuSE 8.2. I don't think it is an OS problem. I have been back in touch with the computer memory geek at Fry's, where I bought the board and he is stumped at this point. There is no newer BIOS on the Gigabyte website than the release version I have. I have not been back in touch with Gigabyte tech support about this yet as I have been pressed for time: I needed a working system first, "extra" come later. On the other hand, that was supposed to be feature of the board, not an extra. I cannot say at this point whether the problem is hardware or BIOS or somehow software based. Summary: board works and seems stable (except for the dual-channel memory issue, above) but SuSE'ers will have to do some legwork and building of their own for drivers. I suspect the raid function should be considered non-supported and not available to SuSE'ers at this point (kernel issue). Hope this helps. I am nothing if not wordy. So my son says. Richard