Re: persistent storage hardware: recommendations, comments, and opinions please
Thanks all for your comments! I'll consider them as I go back through
the available hardware.
On Mon, Jan 21, 2002 at 09:02:10PM -0800, nate wrote:
| <quote who="dman">
|
| > Western Digital WD400BBRTL
| > 40GB, 7200 rpm, Ultra ATA 100, 8.9 ms seek time, 2MB buffer,
| > $130
| >
| > Western Digital WD180ABRTL-120
| > 18GB, 5400 rpm, Ultra ATA 100, 12.0 ms seek time, 2MB buffer,
| > $80
| >
| > Samsung SV4002H (looks like a used disk)
| > 40GB, 54000 rpm, ATA 100, $110
| >
| > The one shop also had 2 Maxtor disks, but I'm not sure I want
| > another one of them.
|
| maxtor is currently my drive of choice. i had some bad
| experiences a few years ago but that was ages ago.
My current Maxtor drive is about 3 years old. I got it to expand my
previous machine, a Compaq (don't get one of those!). The compaq's
BIOS could only find the disk if it was in cable select mode. Thus it
couldn't share a bus with other disks (I don't have the special cable
for it). Later on in its life I had some problems with lost
interrupts. It turned out (after lots of testing and totally wiping
out all data) that the problem was the 110W power supply in the compaq
couldn't really handle all the disks. My dad has a 1.5GB maxtor drive
in the 486 he still has. That disk is many years old, and still
kicking. Back when I was having issues with my disk someone (Luca,
IIRC) said they had seen many problems with maxtors and would never
get one.
If you ask enough people, I think you'll find that all hardware sucks
:-). (makes sense too, since humans are imperfect the things they
make will be imperfect and that sucks)
I'll take a second look at those maxtor drives CompUSA had.
| they
| have much improved since. the opposite of IBM ..who
| was once really reliable and now is complete shit, at
| least for IDE(and yes i have joined the classaction
| lawsuit against them for 12 failed drives in 4 months)
|
| currently, i have 19 80GB maxtor drives in hardware raid arrays,
| 12 of them in raid10, and 5 in raid5, 2 in raid1
| I have another maxtor 100GB not running raid. All
| of these drives are pretty new(less then 6 months old).
| No problems yet.. they are quite reliable sofar.
| Not so fast(i think only 5400rpm..) but fast enough
| for storage. I replaced 15 75GXP 75GB IBM drives with
| the maxtors because of excessive failures on the IBMs.
|
| I don't have experience with samsung though i wouldn't
| personally buy them.
|
| western digital has a rep for being somewhat reliable
| but slow ..
|
| vendors i would purchase for IDE:
| 1) maxtor
| 2) seagate
| 3) ..would keep shopping till i found one of the above..
I do recall seeing Seagate as one of the big players a while back.
I'll add WD to your list since you say they are reliable, and I too
thought they were a good name to go with. (actually, the 486 I have
acting as a desk-weight (no paper under it) has a really old 300MB WD
drive in it that I don't see any problems with)
| > Can someone provide a comparison of Ultra ATA 100 and SCSI? Which
| > is faster and/or more reliable?
|
| depends what SCSI.
Ahh, I see. Time for more research :-). I have been under the
impression that SCSI is better; thought the data rates advertised on
the controllers I found seemed to say the opposite.
| My desktop at work has a single 9.1GB
| Ultra160 SCSI hooked to an Adaptec 29160N Ultra160 SCSI
| card. I also have an IBM 75GXP(the horror!) 46GB drive
| hooked to a Promise ATA/100 IDE card. both are fast, but
| the SCSI(which is IBM too) is CLEARLY much faster. it
| just SCREAMS.
Cool. Better than the 13MB/s my current IDE disk gets now?
How much better? (I haven't done any tuning of it yet, and I don't
even know if it is ATA/100 or ATA/66 or something else altogether)
Digging out the manual, my Gigabyte mobo has an AMD 756 IDE controller
with PIO, Bus Master and UDMA 33/ATA 66 capability.
| I would take out the IBM IDE drive now
| that ive had problems with the others but i don't want
| to lose my 249 day 4 hours uptime. The OS is installed
| on the SCSI, the IDE is used for data only. I don't
| want to put high I/O on the IDE as i am certain it will
| fail.
Sweet. I can't get much uptime because my dad doesn't want me leaving
the machine on all the time.
| Ultra2 scsi i would still take over any IDE.
| Ultra wide scsi is quite a bit slower, but IMO still
| more reliable then IDE in the long run. I would
| use ultra wide scsi over IDE as well. Only thing
| holding me back from doing it is scsi drives are
| still pricy for anything over 9GB.
|
| anything less is worthless for HDs. SCSI2 is fine for
| CDROMs and CDR/CDRW though.
|
| > Adaptec SCSI Card 2906
| > 7 devices, non-bootable, 10 MBps transfer, $63
| > SCSI 1, SCSI 2, Fast SCSI 2
|
| no good, at LEAST 2940UW(40Mbps) or better
Ok, thanks for the tip.
| ..SCSI cards (esp adaptec) are quite expensive.
I thought so, but the above one isn't too high
| the ultra160 i have still goes for at least $200.
ahh.
| though the 2940UW should
| be much less. They last much longer too. I am shocked
| i am still using the 2940UW i got in a system in 1996.
| Everything else from that system has been replaced 4-5
| times over, not the scsi card though ! Note if you
| get a older scsi card like the UW you'll have to be
| sure your drives can run in narrow mode/SE mode(or both?)
| otherwise you'll get massive scsi errors. most drives
| can do this(i can't think of any that can't). but
| keep it in mind if you decide to go scsi. diamond
| used to make a SCSI card line, i forgot the name of
| it, but it had a real good rep(i used their cards in
| a few win32 machines for tape drives, worked good)
| they were REALLY cheap. but they were also discontinued
| long ago, not sure how good modern linux support is on
| them or if you could even find a place that had one.
| For most tasks i would happily use those diamond
| scsi cards over the adaptecs. the diamonds topped out
| at 40Mbps(UltraWide).
If that is bits, then that equates to 5MBps which is half of what the
SCSI Ultra2 card claims to do. Is that correct, or did you mean
bytes?
| > Adaptec ATA RAID 1200A
| > 2 ATA/100 channels, RAID 0 1 0/1 JBOD, bootable, $100
|
| i wouldn't use it, driver support for linux isn't that
| hot. last i heard binary only drivers for distros like
| redhat.
binary only ... that's enough to put my wallet away. I've heard about
RAID before (brief discussion in a class once) and thought it would be
interesting to consider.
| If you want good IDE controllers get Promise,
| their ATA/66 and ATA/100 cards are awesome.
I forgot to mention that the one shop also has a
Promise Ultra 100 Tx2
UDMA 5, 4, 3, 2, 1, 0
$40
(Is UDMA == Ultra ATA, or what's the relationship between the terms?
Gotta look it up ... where's google again ;-)?)
| I would not
| buy the "raid" editions though. If you want real hardware
| IDE raid go with 3ware. though i could tell some good
| nightmares about 3ware ide raid..even after all the trouble
| i'd still choose them over any other IDE raid vendor soley
| due to better linux driver support. im waiting a good
| 6 months while i run my 4 3ware systems at the office
| thorugh tests before i consider using 3ware at home
| after the problems ive had :)
I remember your bug report. Sounds like RAID isn't "all that" in
practice, though in theory it looks good.
| > CompUSA (Silicon Image Sil0649CL160)
| > Ultra ATA 100, 2 channel, ACPI, $30
|
| don't know ..but if its not a promise, i wouldn't use it
| in my system(BTW promise ATA/100 is around $35)
Like the one I forgot to list? :-)
| > Does RAID restrict the combination of disks I can have? IIRC RAID
| > 0 is no redundancy, and RAID 1 is simply maintaining two copies on
| > separate disks. If so, then wouldn't both disks need to be the
| > same size?
|
| optimally both drives should be exactly the same, same model
| same brand, and same speed. but if they are different, the
| raid array will just be as big as the smallest drive. i setup
| a software raid5 array on 4x9.1GB ultra2 scsi drives accross
| dual 2940UW scsi cards, the drives were slightly different,
| 2 were identical IBM, 1 was a faster bigger IBM, the last was
| a quantum ..works good though.
Cool. Given your warnings (and the $$ attached to it) I'll probably
only go as far as trying software RAID, if anything.
| a warning for software raid5.. ive only setup 1 raid5
| array(raid 0.96??) but from what i could see you could
| not partition the array up in any way or setup multiple
| raid5 arrays accross partitions(i do this often in raid0
| and raid1). because of that i couldn't get a swap partition
| on the disks with the raid array so i have a 4.5GB disk
| in there for swap. I am able to boot off the array though.
| I am forced to use ext2 as the filesystem as reiserfs
| has severe problems under software raid5/kernel 2.2(i don't
| use kernel 2.4 so i don't know if its any different ..)
| not a big deal as i reboot once or twice a year at
| best ..
|
|
| I run Dual 40GB raid1(software) in my main desktop at home with IBM
| drives(again, bought them before i suffered the rash
| of failures).
|
| my main fileserver at home has 2 x 40GB IBM drives, 1 x 46GB
| IBM drive and 1 x 100GB maxtor drive, hooked to a promise
| ata/100 card. another web/ftp server at home has a 4GB and
| a 6GB drive(maxtors)
|
| one big fat warning about Promise IDE controllers. i have
| found on at least 4 different systems the IDE controllers
| did NOT LIKE any cable longer then 18"(technically anything
| longer violates the IDE spec..) and i get non-critical CRC
| errors during high I/O. Enough to make me switch back to
| 18" cables. Some other controllers like the 3ware raid
| has no visible problems with using cables as long as 36"
| (40/80wire)
Interesting. I'll keep this in mind.
| hope this helps
It does (this goes for everyone). I like hearing from real people who
have worked with the stuff better than reading marketing handouts.
The latter isn't interested in me at all and tends to try and confuse
the reader with buzzwords and/or lack of real information.
-D
--
The crucible for silver and the furnace for gold,
but the Lord tests the heart.
Proverbs 17:3
Reply to: