Hm , I have not the possibilty to set up a hw-raid and mostly not the
time to do so in order for testing and comparisons, but I have a
suggestion towards hw- and sw-raids.
I am using sw-raids on systems like you described with just more RAM and
controllers on it running linux-2.0.36 (will switch to 2.2.x next days).
We also got workstations and servers from HP with connected raid-systems
(raid5). The systems are almost machines with two PA-RISC processor and
1GB RAM and above. The speed of the so called disk-arrays on the HP
machines is very low. Based on UW-disk the only produce a "throughput"
at roundabout 5 MB/s.
The cause seems to be the integrated raid-controller, which works at 40
or 50 MHz and calculates the checksum of the incoming data.
The sw-raid of the linux-machines I use is much more faster even on a
P166 or P200Pro (single processor) and of course on the double PII-400 -
the lx-machines cost a fraction of the HP and redundancy is also given;
also, the hw-raid controllers for Intel based PCs are much more
expensive than normal SCSI-controllers and I think they are also slower
than the sw-raid because of the CPU-speed which is responsible for the
speed of the sw-raid.
This is just a suggestion and a very subjective point of view - I know -
but theory and practice are very close.
Greetings, Dietmar
Francisco Jose Montilla wrote:
>
> Hi,
>
> On Wed, 14 Apr 1999 [EMAIL PROTECTED] wrote:
>
> > On Wed, Apr 14, 1999 at 02:55:21PM +0100, Piete Brooks wrote:
>
> > > I'm a fan of SW RAID (a motherboard with 3 LVD busses, 2 EIDE busses,
> > and 1 W > bus kind of satisfies our disk IO requirements), but I've
> > being pressured to > buy some HW RAID for our "main" NFS servers.
>
> > What do you need the lower performance for ? :)
>
> > Seriously, if anyone can submit a benchmark (bonnie & hdparm -t, or
> > others of your choices), that shows a HW RAID that runs faster than SW
> > RAID, on the same disks, CPU, memory, board, etc, I'd be very happy to
> > see it.
>
> > I've seen a few examples on this list that showed SW RAID running
> > several _times_ faster than HW RAID. But I have yet to see HW running
> > faster than SW.
>
> > I'm sure it can be done. But I'd like to see it.
>
> It happens I'm going to have the chance of doing this kind of
> tests. Please post what kind of tests (apart from bonnie) would you like
> to see, different raid levels, setups, etc.
>
> I'm going to set up a machine with RH 5.2 and 2.2.5 with raidtools
> 0.90 and appropiate kernel patch (I merged the 2.2.3 one to be applied
> to a 2.2.5).
>
> Here's the configuration of the testbed machine (SMP) I'll setup:
>
> Asus P2B-DS (Dual mb with integrated AIC-7890 Ultra 2 SCSI)
> 2x PII 400
> 2x 128 PC100 non-ECC DIMMs
> Kingston KNE-100TX nic.
>
> I think this is a pretty affordable production machine, so no one
> could say later that it was built in order to enhance sw raid.
>
> System will be on a non-raid partition of a "system disk" hooked to the UW
> bus, easing swaping raids for testing.
>
> The hw raid will be a Mylex DACPGM-2-8E with external cabinet.
>
> I'll have 8 4,5 Gb U2W discs and 2 9,1GB U2W discs.
>
> Of course, these aren't for me :) I only will have everything for
> a couple of days, that's we'd better have all the tests in a schedule
> list.
>
> I hope to build a fair and serious comparison table between sw/hw
> raid with the help of the list, including both "raw" benchs (eg. bonnie,
> iozone) and application tests (i.e., nfs, etc...) but i'll need the
> collaboration and suggestions of all people interested.
>
> Greetings,
>
> *****---(*)---**********************************************---------->
> Francisco J. Montilla Systems & Network administrator
> [EMAIL PROTECTED] irc: pukka Seville Spain
> INSFLUG (LiNUX) Coordinator. www.insflug.org - ftp.insflug.org
--
We all have lack of knowledge...
Dietmar Stein, Systemadministrator UNIX/Linux
[EMAIL PROTECTED], [EMAIL PROTECTED]
