• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

first RAID setup... any pointers?

software RAID in Windows means dynamic disks
a whole lexicon of special considerations youll need to learn
, like a bearly understandable dialect of a language you speak


Dynamic Disks
Description of Disk Groups in Windows Disk Management
Dynamic vs. Basic Storage in Windows 2000
Basic and Dynamic Disks @ Windows & .net Magazine
HOW TO: Recover an Accidentally Deleted NTFS or FAT32 Dynamic Volume
Dynamic Disk Hardware Limitations (No firewire, USB, removable or laptop)
HOW TO: Set Up Fault-Tolerant Sets on Dynamic Disks in Windows 2000 <<<<<<<<<<<<<<<
Dynamic Disk Numbering and the DmDiag.exe Tool
HOW TO: Regenerate a Dynamic Mirrored Volume in Windows 2000
Restrictions on Extending or Spanning Simple Volumes on Dynamic Disks

Limits of Dynamic Disks in Windows 2000
LDMDump (Freeware utility) @ sysinternals
LDM Database @ Linux-NTFS project
LDM FAQ @ Linux-NTFS project


in addition they come with a considerable performance penalty since the CPU needs to do all the parity calculations, wheras a good RAID card will have its own processor or engine and its own oncard buffer\RAM

one last point
once you convert to a dynamic disk, you can reconvert it back and retain the data without shelling out the $$$$ for an ap that will hopefully do it correctly
 
Everfalling said:
you mean that you dont think there's any more performace from having 4 drives spin to access a single file, rather than one drive spin for that same file? my assumption is that 4 drives doing the work of 1 is faster, i'm guessing i'm wrong then?

ok, so if raid 0 is too risky just to have one big TB drive, is there another way to do this? is there another way to make as if all 4 drives are one big drive? or would i be stuck managing 4 seperate ones?

Read through this forum and look at the benchmarks. 4 drive raid 0 is not faster than a single drives in most desktop applications
ribs
 
ribs1 said:
Read through this forum and look at the benchmarks. 4 drive raid 0 is not faster than a single drives in most desktop applications
ribs
yeah, reading these guy's posts so far have told me pretty much just that. now i'm thinking raid 5... hmm... we'll see..i just got the drives in the mail... they'll prolly sit in the box for a little bit till i get this all worked out...
 
Everfalling said:
kinda like this:

1st set of sata ports:
[max] [max]
[max] [max]
2nd set of sata ports:
[max] [WD]
[xxx] [xxx]

where as all the 'max' drives are on Raid5, while the WD is left alone to be it's own seperate C drive.

yes, that would work, however, I would recommend to have the C: drive on the nvidia controller, since you would not need to install SATA drivers during Windows Setup.
 
its also my understanding that
youd see a considerable advantage with a Linux software RAID5
 
Everfalling said:
where as all the 'max' drives are on Raid5, while the WD is left alone to be it's own seperate C drive.
This setup is possible with software raid 5. However, doing it under windows has two pitfalls. First, you'll need to get a server OS - windows server 2003 or 2000 - because MS didn't build raid 5 support into its "consumer-level" OSes. Second, write speeds with WSR are abysmal from all reports.

Linux SR is a good deal less problematic; you can get it with any off-the-shelf distro and writes are pretty good - a single disk's speed or thereabouts with my setup.

 
unhappy_mage said:
...and writes are pretty good - a single disk's speed or thereabouts with my setup.
Not if your hard drives were made on planet Earth. Writing to two disks is always slower than writing to one.
 
masher said:
Not if your hard drives were made on planet Earth. Writing to two disks is always slower than writing to one.
What? I'm talking about STR here, not latency. The STR of my setup is about 50 MB/s reading or writing. Sure, latencies are hurt; I'd hate to load farcry levels from it, for example. But raid would be useless if it *always* hurt performance. A raid 0 stripe set is a good deal faster than a single disk in STR, why should raid 5 be any slower for big-block writes?

 
unhappy_mage said:
why should raid 5 be any slower for big-block writes?

parity calculations without an adeqate buffer

personally I don't notice anything with my SX6000
but thats got 128MB buffer and a dedicated RISC Processor for parity

as to what someone might notice in software
guess it would depend on the CPU(s) load at the time of the write and system memory
 
unhappy_mage said:
What? I'm talking about STR here, not latency.
Well you didn't specify originally...and on a desktop, write latency is more important than STR usually anyway...there aren't many tasks that require large sustained writes. And, as Ice Czar has pointed out, software Raid 5 eats up substantial CPU time for parity calculations...in many cases, enough to cause a bottleneck right there, regardless of disk speeds.

Bottom line is desktop Raid 5 is always slower than a single disk, and software Raid 5 is substantially so.
 
masher said:
At my lab, we have just over 150 Raid-5 arrays, each containing 14 disks, and each running 24 hours/day for the past 7 years. Plenty of disks have failed...but we haven't lost a single byte yet.

The real dangers to Raid 5 arrays are _external_. If you have a software flaw that writes bad data into the array-- you lose the data. If someone nukes your site-- you lose the data. Otherwise, as long as you rebuild as soon as any disk fails, your risks are nearly negligible.



Alright and at my old lab we have 7 full cabinets of drives and we have two raid5 arrays fail in a row. Who knows how many we have now (15k servers ATM) Granted, two out of the x number of years they have been running... but it still counts as a fail.

Anything is possible ;) Just because your lucky and it didn't happen to you yet, doesn't mean it wont happen.
 
Ockie said:
we have two raid5 arrays fail in a row. Anything is possible ;)
Well of course anything is possible, but the only way I've ever seen raid arrays fail is:

- Failed drives not replaced in a timely fashion, so the array ran out of hot spares.
- Controller card failure
- Running non-TLER drives in an array or similar configuration error.

The case of a drive failing because two drives failed near-simultaneously is theoretically possible as well...though astronomically unlikely in practice.
 
unhappy_mage said:
This setup is possible with software raid 5. However, doing it under windows has two pitfalls. First, you'll need to get a server OS - windows server 2003 or 2000 - because MS didn't build raid 5 support into its "consumer-level" OSes. Second, write speeds with WSR are abysmal from all reports.

Linux SR is a good deal less problematic; you can get it with any off-the-shelf distro and writes are pretty good - a single disk's speed or thereabouts with my setup.


It is possible to do software Raid 5 in Windows XP pro. Google it.
 
masher said:
Well you didn't specify originally...and on a desktop, write latency is more important than STR usually anyway...there aren't many tasks that require large sustained writes. And, as Ice Czar has pointed out, software Raid 5 eats up substantial CPU time for parity calculations...in many cases, enough to cause a bottleneck right there, regardless of disk speeds.

Bottom line is desktop Raid 5 is always slower than a single disk, and software Raid 5 is substantially so.
Okay, in latency terms. But your definition of "substantial" and mine are different - I use about half a percent of my CPU doing XORs while writing at full speed. Think about it - with a chip at a gigahertz (approximately), I'm handling 50 million bytes a second. Put them in word-sized chunks (12.5 million) and you're at 1% cpu usage. Do SSE things and you can probably do a bunch in one clock. And if you're doing large file things, latency is a non-issue. All I'm saying is, "not if your disks were made on Earth" is a little extreme, no? A raid array (if not a raid *5* array) is definitely faster in certain situations, or it would have been abandoned a long time ago.

ribs1 said:
It is possible to do software Raid 5 in Windows XP pro. Google it.
The policy of this forum, as far as I recall, is that this is hacking your operating system, and therefore illegal and not a topic of discussion. It's okay to mention that it can be done, but not to say how to do it.

 
unhappy_mage said:
The policy of this forum, as far as I recall, is that this is hacking your operating system, and therefore illegal and not a topic of discussion. It's okay to mention that it can be done, but not to say how to do it.


hmmm guess Id better write myself up :p
I thought it was very much a grey area, being more an advanced tweak than hack

apon consultation it appears as if I was in error

DougLite said:
The code to support software RAID-5 ships in the Dynamic Disk Management code in Windows XP Pro, however it is deliberately disabled. Software RAID-5 support is reserved for server products. The 'hack' that was referred to in that thread involves changing the registered product type, which is a violation of the Windows XP Pro EULA. For more information: http://www.hardforum.com/showpost.php?p=1027594195&postcount=7

and as a violation of the EULA unhappy_mage is indeed correct ;)
 
so since I personally think thats stupid taking away something that had been previously included in W2K pro
not to mention shipping it, but making it illegal to activate.
And since I don't agree that only enterprise servers need data protection and that the end user can't benefit from redundant parity disks

here is a nice legal alternative that should make Redmond happy :p

RAID and Data Protection Solutions for Linux
odds on favorite, available in most full distros
 
would all this softeware and os problems be solved by just buying a decent controller card? or are there still problems with that? cause i'm not gonna switch to linux, and i'm not haxxoring my XP. i think i'll just ask for suggestions in hardware then :)
 
absolutely

http://www.tweakers.net/reviews/557/1
of course alot of those are likely not 32bit 33MHz PCI

just stay away from 64bit 66>133MHz PCI-X varients that arent 32bit compatible
youd need to confirn they are both 32 and 64 bit compatible (which a few are)
you dont have 64bit slots on your board, but some of the cards will fit in 32bit slots anyway

DOH there are no PCIe in that roundup :p



http://www.areca.com.tw/products/html/pciE-sata.htm employs
PCIe x8 which I think? can run with a single PCIe vidcard
it would be choking the vidcard to PCIe x8 too though
(since you have either a single usabl;e x16 or a dual at x8 each)
I dont really see any PCIe x1 RAID5 Cards from the limited searching Ive done

there are lots of 32bit 33MHz PCI RAID5 cards from Highpoint, Promise, 3ware, Adaptec, ect
but most arent SATA, there are a few, so most likely your back to 64 bit SATA cards that are 32bit compatible
however you dont see many performance benchmarks of such cards throttled down to their 32bit performance
 
as a note, its unlikely youd ever need the 64bit bandwidth anyway since its damn near impossible for a single user thats not doing say realtime video editing or something to saturate the 32bit 33MHz bus anyway (133MB/s actual sustained throughput to the drives) where that often changes is in a multiuser environment, NAS or SAN

cut and paste 101

some bus numbers
32bit 33MHz PCI Bus 133MB/s, SATA port not attached to the normal 133MB/s southbridge 150MB/s
32bit 66MHz & 64bit 33MHz PCI Bus 264MB/s, 64bit 66MHz PCI Bus 508MB/s,
64bit 100MHz PCI-X Bus 800MB/s, 64bit 133MHz PCI-X Bus 1066MB/s
and past that the various PCI-Express lanes at 250 MB/s per x2, x4, x8, x12, x16 and x32 (3GIO 3rd generation I\O) single direction

of course traffic on the bus from other sources factors in, CPU(s) overhead, or lack there of on the controller(s) ect, would impact the throughput as would the boards chipset achitecture.
All additionally constrained by the controller(s) in question and the internal limitations of the drives themselves
As the Disk Spins
and the stripe width of the array (number of drives), with additional considerations of stripe size to typical access\file
 
Limitations of PCIe

Bus.................Bandwidth Single Direction.....................Bandwidth Dual Directions

PCIe x1..............................250 MB/s.............................................500MB/s

PCIe x2..............................500MB/s.............................................1000MB/s
.
PCIe x4.............................1000MB/s............................................2000MB/s

PCIe x8.............................2000MB/s............................................4000MB/s

PCIe x12...........................3000MB/s............................................6000MB/s

PCIe x16...........................4000MB/s............................................8000MB/s
 
but my mobo only has 3 PCI, 2 PCIex1, and 2 PCIex16

is PCIex8 just a dummed down version of x16? like, same phisical slot size but less power?

cause right now, i have 2 of the pci's open, both pciex1's, and one pcix16. i'm also looking at cost. if a regular pci card runs a little slower than one on a pciex16, but it's cheaper, i'd go with the regular pci, cause it's just going to be managing movies and music.
 
you can put anything smaller into a bigger slot in PCIe
so yes that x8 RAID card would go into x16

and I actually found a few folks running it on that board with a Quadro PCIe

as far as cost, the thing is that RAID5 isnt all that common with end users especially in SATA, so there isnt alot of 32bit 33MHz cards out there that will do that level
in enterprise cards there are but they get pricey, you have PCIe x8 options, and lots of 64bit cards that you need to make sure are compatible with 32bit

Tyanslot.jpg


32bit 33MHz on the left
64bit PCIX on the right

some cards will fit in either, with the extra fingers just hanging off the smaller slot
for instance here is a 32 or 64 bit compatible SCSI card
http://www.adaptec.com/worldwide/pr.../SCSI+(Premium)/Ultra160+SCSI+PCI+HBAs+&+RAID
 
Ice Czar said:
you can put anything smaller into a bigger slot in PCIe
so yes that x8 RAID card would go into x16

and I actually found a few folks running it on that board with a Quadro PCIe

as far as cost, the thing is that RAID5 isnt all that common with end users especially in SATA, so there isnt alot of 32bit 33MHz cards out there that will do that level
in enterprise cards there are but they get pricey, you have PCIe x8 options, and lots of 64bit cards that you need to make sure are compatible with 32bit

so what are you saying? the only thing i'd find are mostly 64 bit cards, and if i found one for 32 i'd be expensive? ok, so my challange would either be, finding a cheap 32 bit, or finding a cheap 64 bit that's 32 bit compatable?
 
masher said:
Well of course anything is possible, but the only way I've ever seen raid arrays fail is:

- Failed drives not replaced in a timely fashion, so the array ran out of hot spares.
- Controller card failure
- Running non-TLER drives in an array or similar configuration error.

The case of a drive failing because two drives failed near-simultaneously is theoretically possible as well...though astronomically unlikely in practice.


Replaced right away, was actually during testing when we were pulling drives making sure they rebuild.

Controller cards worked fine, was just a hickup or something

Drives were cheetah scsi's
 
I'm starting to think that this guy is going to get overinvested here. He was looking to find a way to use the 4 drives he had, and now he's considering getting another drive and a controller that will cost a couple hundred dollars?

What ever happened to JBOD.
 
TheTMan said:
I'm starting to think that this guy is going to get overinvested here. He was looking to find a way to use the 4 drives he had, and now he's considering getting another drive and a controller that will cost a couple hundred dollars?

What ever happened to JBOD.

'ok... so... RAID 0 is bad for small files, doesn't increase proformance unless for larger files, has a chance of failing like, to the power of 4 more than usual (or some other big number) and if it does fail, trags down everything with it? hmm... and JBOD doesn't do much better, though if one drive fails, it might just take down just its own data, but it'll still screw things up, and evidently copying stuff back and forth is a hassle... well shoot.. so much for my day dreams of one large terabyte drive >>'

that's why no JBOD. i'm still thinking about it, keeping my options open. theoretically, the 4 drives i bought wont even need to be paid for for another 6 months thanks to the newegg preferred buyers card, so taking an actual chunk away for one more drive and a controller isn't so bad, even though in total i'm prolly gonna end up spending about a grand if i do it this way (5 drives + controller). but at least i'll have the error protection of Raid5 to back it up, rather than completely lose an entire drive, and slightly screw up the others using JBOD. or am i still going about this wrong? :/
 
Everfalling said:
'ok... so... RAID 0 is bad for small files, doesn't increase proformance unless for larger files, has a chance of failing like, to the power of 4 more than usual (or some other big number) and if it does fail, trags down everything with it? hmm... and JBOD doesn't do much better, though if one drive fails, it might just take down just its own data, but it'll still screw things up, and evidently copying stuff back and forth is a hassle... well shoot.. so much for my day dreams of one large terabyte drive >>'

that's why no JBOD. i'm still thinking about it, keeping my options open. theoretically, the 4 drives i bought wont even need to be paid for for another 6 months thanks to the newegg preferred buyers card, so taking an actual chunk away for one more drive and a controller isn't so bad, even though in total i'm prolly gonna end up spending about a grand if i do it this way (5 drives + controller). but at least i'll have the error protection of Raid5 to back it up, rather than completely lose an entire drive, and slightly screw up the others using JBOD. or am i still going about this wrong? :/

Nah, you are thinking it through well. JBOD would be significantly cheaper than raid 5, with no redundancy.....

But RAID5 is just so damn cool. ;) I want to do it too, but I'm broke as a joke.
 
Everfalling said:
so what are you saying? the only thing i'd find are mostly 64 bit cards, and if i found one for 32 i'd be expensive? ok, so my challange would either be, finding a cheap 32 bit, or finding a cheap 64 bit that's 32 bit compatable?

that I dont see alot of 32bit SATA RAID cards capable of RAID 5
and since SATA is rather new in comparision to IDE or SCSI.you dont see them much on ebay or cheap from the manufacturers.

as a storage array you dont really need alot of peformance
but having at least the potential to survive a single HDD failure is nice
not a substitute for backup though, any number of issues can still loose your data from pilot error, to filesystem corruption, power events that kill more than one drive and of course malware.

Ive been running a hardware RAID 5 for some 5 years now (IDE) and its certainly an investment considering that the efficiency of a RAID 5 is determined by its stripe width
total storage equaling the additive of the number matched drives minus one drive
3x100GB=200GB, 6x100GB=500, 12x100GB=1100GB

back to PCIe, 64bit and 32bit
SATA as a storage alternative to SCSI in the enterprise has been widely adopted, but enterprise storage generally has 64bit mobos, so unlike SCSI and IDE there arent alot of 32bit compatible cards, the enterprise has moved past that era, there are plenty that do the lower levels of RAID but not that many that do parity levels, there are quite a few PCIe since that is the future but again they are expensive.

you need to hunt for something that is going to meet your needs and budget and is 32bit compatible or your going to have to pop for a PCIe and research compatibility with your choosen vidcard, I tried to do a little of that, but Im not sure what kind of bandwidth impact it might have on say a gaming box, though actual compatibility looks good Quadros are workstation vidcardsbased on the same chips that gaming cards are.

my personal feeling on JBOD is why bother? there is no redundancy, it complicates data recovery if a drive fails, there are very limited reasons to have dynamically expandable volumes. Better to just have properly backed up individual disks, and if one goes so be it.

as far as software RAID there are some advantages for redundancy, and penalties for performance, but we dont know what your performance needs really are, you may have plenty of resources to spare you may not, youd have to be writing alot of data to impact a modern CPU substantially and doing something pretty CPU intensive things concurrently
but its either hack the OS you have, or switch OSs
 
well crud. that woulda been awsome too. it's only 78 bucks on ebay right now with a bit more than a day left... well this is going to be annoying...
 
watch it on ebay
thing is, everyone would need to be asleep for it to stay at that price
I buy alot of stuff on ebay. and there is absolutely no point bidding on items till the last seconds of close, Ive seen items go from $25 to over $750 inside of 2 minutes
 
well, there's no point watching it anymore now, you just said it would be compatable with my computer... no point buying it now...right?
 
ok... i forgot about this, but i'm not sure it's right. is it possible to daisy chain SATA together? even if it's only 2 drives per cable like IDE? cause if so, wouldn't that solve my problem?

EDIT:
ok, i think i might have found my own answer. i;m thinking instead of getting one big controller for 5+ sata150 ports, i'm gonna get two controllers, each with, lets say 4 sata150 ports. my results come out way better, esp considering they use regular PCI slots. tell me what you think:

http://www.newegg.com/Product/Product.asp?Item=N82E16816102065 < SATA II 4-port controller

i'm assuming SATA includes sata150. if not i have these choices:

http://www.newegg.com/Product/Produ...50410+1193512552+1193212547+1193412759&ATTR5=

SATA 4-port controllers. I'm, again, assuming sata150 fits within the relm of regular sata.
 
ok, well poop. i just refinded the search. seems there are only two cards with the ability to do raid 5 on them, and they're both like 200-300 bucks.... hmm.... but one of the feedbacks on the Promise card (first link) says he used a software raid... now..considering all i'll be doing is storage, and considering he said he got it to work on his xp pro, would cpu usage really be an issue? take into account i'm running a 3800+ amd in this thing. cause if the sfotware raid isn't a problem, the promise card is basically exactly what i need right? i just gotta buy two of them to fit the 5 drives. oh please say it'll work >.<
 
yeah... feel kinda bad about bumping this again but i still need help finding a RAID controller that'll support what i have... money isn't even a problem as long as i know there are things out there that do what i need... thanks if anyone can help
 
Back
Top