Call to anyone with RAID 1E experience

[Digital Technophile]

Howdy,

I've been doing some research into RAID controller card in preparation of a new PC build I'm planning. For a long time I was really sold on RAID-5, until someone pointed out the drawbacks which I agree outweigh the pro's of that mode and started reading up on RAID-6. But I recently found this Areca card, which fits my budget, satisfies my requirement for PCIe bus, and supports a whopping 12 SATA II HDDs! Looking over the specs, I notice a RAID mode I've never seen before... RAID 1E (E=enhanced), which I understand is an IBM developed RAID mode.

Quiz time:
1) Anyone have experience with this Areca 1231ML card?
2) Does anyone have experience with RAID-1E?
3) If question 2 is yes AND you have experience with RAID-6, which would you prefer for a 4-6 drive array and why?

I'd also just like to point out this card supports both modes (1E and 6), so unless anyone has some definite reasons I shouldn't get this card I'd like to discuss the pros and cons of these RAID modes.

Thanks!

[Uranium-235]

Re: Call to anyone with RAID 1E experience

raid 5 (single redundant) and 6 (dual redundant) still suck donkey balls

go with raid 10, decent size and less of the parity complications

raid 1e is non standard, and google is wonderful
http://www.techrepublic.com/article/...aid-1e/6181460

[mockingbird]

Re: Call to anyone with RAID 1E experience

Are you using Windows or Linux? What type of environment will the computer sit in? I would stay away from proprietary stuff.

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

Use RAID-1.
If you need speed use 10,000 RPM drives.

Generally only a small portion of data needs that much speed anyway.
The files that aren't accessed often can be on a slower array.

To get a little more performance use [say] a 36 or 74 GB 10,000 RPM Raptor [or some SCSI 10k] as a dedicated swap/page file drive.
It doesn't need to be mirrored so can be on a mobo connector.
[Speeds things up by freeing the drives on the array of the Virtual Memory and/or Swap File duties completely. The heads have more time for the 'real' job.]
.

[Digital Technophile]

Re: Call to anyone with RAID 1E experience

raid 5 (single redundant) and 6 (dual redundant) still suck donkey balls

go with raid 10, decent size and less of the parity complications

raid 1e is non standard, and google is wonderful
http://www.techrepublic.com/article/...aid-1e/6181460

Hey that's the article I found when I first searched to see what it was. Nested 1+0 (stripe of mirrors) seemed like a better option than RAID-5, although the latter seems to be more tolerant of a bunch of different sized drives. RAID-6 is a new one to me, but 1E sounds like the most attractive mode to use.

@mockingbird: I've still got my copy of Server 2003 Standard edition, which I want to get going again (not installed ATM, don't get me started...long story), but I've got an interview with GE Aviation on Tuesday and IF it goes well I'd really really like to put something together that runs VMware ESX; got those hungry eyes for that MSI 890FXA-GD70 main board right now possible paired to a 6-core processor, who knows?

@PCBONEZ: I'm still working on (got so many damn irons in the fire ATM) getting my Intel D850EMV2 MoBo recapped and upgraded to pc1066 rambus. I also found on fleabay (GOD I love to hate those scumbags) some still sealed 3.5 10k Raptors (WD1600ADFS) that I'm also watching and drooling over. Any recommendations for a PCI bus SATA adapter; Server 2k3 + Intel D850 + SATA II PCI RAID card + 2 WD1600ADFS (80GB) in RAID-1 sounds like good recipe for new domain controller, eh?

But back to RAID-1E. That Areca 1231ML card supports many RAID modes, but the thought of migrating between different modes sounds time consuming to say the least. Trying to do mathmagic here is making my head hurt. Between 5, 6, and 1E, which has the greatest parity overhead? From what I can gather It's RAID-6, but let's through some variables in the mix.

What about 4 disk arrays from the three modes in question versus 6 disks? Or even 8 disks (man that better be one honking powersupply; err, maybe if we're talking fault tolerance better make it two honking PS's ), Would there be a difference in the amount of overhead; it seems like an array of more than 5 disks and parity becomes more important to justify the overhead.

Or how about the size of the disks, which mode become more efficient - an array of 80GB hdds, or one made up of 160 or even 320GB. Then we have to think of time to recover if one drive fails.

And let's just for the sake of argument consider a 6 disk nested Raid 1+0; three striped mirrors, still a 50% overhead even if it is not overhead associated with parity. Is it faster than the three aforementioned modes?

And after reading through IBMs pages on RAID levels, I've become curious of the difference between "data redundancy" (RAID-1,1E, 1+0) and "data protection" (RAID-5 and 6). Thoughts? Oh and they really threw a curve ball when they mentioned the existence of modes like 1E+0, 5+1, etc.

Man, that's a lot to consider, but I've got time. I want to work through this multitude of RAID modes to figure out what has the best performance while maintaining redundancy (or protection?); I haven't checked yet, but I hope that Areca card will work with ESX server...that's going to be the end result of this whole RAID discussion as far as I'm concerned. I'm still rethinking the whole of my storage needs as well; my data's still held hostage by that SATA III and my own blind ignorance and will need a safe place to go when I get it back. But I digress (probably because the hour grows late) for now.

As always I value your C & C - thanks!

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

Any recommendations for a PCI bus SATA adapter....

I'm assuming you mean a RAID SATA controller.
For 5v PCI slots I [now] use 3Ware 9500 series.
They are PCI-X but they work in standard PCI slots when there is clearance in front of the slot for the card extension.
-- Use some electric tape over the otherwise exposed fingers. [Connectors]
Some people just hack off the extension.
I haven't gone that far. [Yet.] Haven't needed to.

If you have a board with the rare 3.3v PCI then you can use the 3Ware 9550 series.
The 9500 series is 3.3v -or- 5v.
The 9550 series is 3.3v only.

Pay attention when shopping for those though.
Some of them use multi-lane connectors vice standard cables.
The special cables aren't hard to get but it's a bummer when you think you are about done and discover you don't have the right cables.
.

[Uranium-235]

Re: Call to anyone with RAID 1E experience

the multi-lane SFF 8087? You can get fan-out sata cables for those

I would imagine raid 6 has the most overhead cause it has to calculate two drive parity stripes per drive

the size of raid 5 is (n = number of drives, s = size of drives)
s * (n - 1) for raid 5
s * (n - 2) for raid 6

nested raid 10 IMO is till the best bet cause it gives you good space and performance

raid 5 is still ok if you have a lower number of drives, cause the more drives you have the higher chance of rebulding while another dies, in the case of raid 6 chance of 2 dying, but it has big overhead

raid 10 dosen't have much overhead cause there is no parity, only striping

[Digital Technophile]

Re: Call to anyone with RAID 1E experience

I'm assuming you mean a RAID SATA controller.
For 5v PCI slots I [now] use 3Ware 9500 series.
They are PCI-X but they work in standard PCI slots when there is clearance in front of the slot for the card extension.
-- Use some electric tape over the otherwise exposed fingers. [Connectors]
Some people just hack off the extension.
I haven't gone that far. [Yet.] Haven't needed to.

If you have a board with the rare 3.3v PCI then you can use the 3Ware 9550 series.
The 9500 series is 3.3v -or- 5v.
The 9550 series is 3.3v only.

Pay attention when shopping for those though.
Some of them use multi-lane connectors vice standard cables.
The special cables aren't hard to get but it's a bummer when you think you are about done and discover you don't have the right cables.
.

Thanks PCBONEZ, I'll check that card out. The Intel board I have has 5v PCI 2.2 slots (and slot 4 has SMBUS routing enabled), so I'll take a look at the 9500. I'm sure I'll have room for PCI-X using your suggestions; never considered doing that and to be honest never read up completely on PCI-X bus specs. Getting this board going as a domain controller just going to require pc1066 RDRAM (seems to becoming more affordable finally), the card and two drives (thinking WD1600ADFS Raptors). I've got two boards the first of which blew two Nichicon HM 6.3v 3300uF 105c caps (oddly the only two Nichi's on the board; the balance are United Chemicons and some polys on the vcore). The second one was still in service when I took it offline, but I'd like to try recapping both starting with the earlier rev. board first. More on that coming soon (some interesting oddities regarding the selection of caps on the earlier board vs. the later board).

the multi-lane SFF 8087? You can get fan-out sata cables for those

I would imagine raid 6 has the most overhead cause it has to calculate two drive parity stripes per drive

the size of raid 5 is (n = number of drives, s = size of drives)
s * (n - 1) for raid 5
s * (n - 2) for raid 6

nested raid 10 IMO is till the best bet cause it gives you good space and performance

raid 5 is still ok if you have a lower number of drives, cause the more drives you have the higher chance of rebulding while another dies, in the case of raid 6 chance of 2 dying, but it has big overhead

raid 10 dosen't have much overhead cause there is no parity, only striping

Thanks for all the RAID tips, they are certainly appreciated. I want to get a definite idea of which mode I want to use when the time comes to implement that array. Still somewhat confused over the terms "data redundancy" (mirroring) vs. "data protection" (parity).
It seems that redundancy means that the data is simply mirrored with no error correction, whereas protection (parity, like with Raid 5 & 6) seems to imply some error correction (like ECC RAM does).
And what I've come to understand about Raid-1E, it simply means that striping and mirroring occur on a single unnested level.
So really what I'm down to now and again state the purpose of exploring these modes is for VMware ESX server - will the overhead for RAID 5/6 be that much of a performance penalty considering what that overhead provides ( redundancy ≠ protection (error correction please correct me if I'm wrong). Worse comes to worse, I'll get the components together for this upcoming build (the Areca card supports ALL of these RAID modes we've been discussing) and test the performance of them.

And URANIUM-235 nested RAID-1+0 is probably the "safe" route. But I don't consider sucking donkey balls a reasonable enough argument not to use them; I'm the kind of guy who also wants to know why (yes mommy hated me when I was a kid ).

Cheers

[gerg]

Re: Call to anyone with RAID 1E experience

The 9550 series are solid cards I would stay away from the 9650 series, I have 100 of them in service and the cards are beginning to fail after only being in service for a year. I have had 3 begin to fail in the last 30 days and it sounds like a common problem with the controller card. It will randomly drop my second raid 1 drives in my suse VM boxes.

I'm assuming you mean a RAID SATA controller.
For 5v PCI slots I [now] use 3Ware 9500 series.
They are PCI-X but they work in standard PCI slots when there is clearance in front of the slot for the card extension.
-- Use some electric tape over the otherwise exposed fingers. [Connectors]
Some people just hack off the extension.
I haven't gone that far. [Yet.] Haven't needed to.

If you have a board with the rare 3.3v PCI then you can use the 3Ware 9550 series.
The 9500 series is 3.3v -or- 5v.
The 9550 series is 3.3v only.

Pay attention when shopping for those though.
Some of them use multi-lane connectors vice standard cables.
The special cables aren't hard to get but it's a bummer when you think you are about done and discover you don't have the right cables.
.

[Per Hansson]

Re: Call to anyone with RAID 1E experience

I don't have experience with that particular RAID mode but I would stay away from it because there are many properiatry modes like that, from different manufacturers
Like RAID 50 etc etc

What they all share in common is that if you get in trouble you really are gonna have a hard time to fix it, since it's such a special setup there wont be much info on the web about recovery procedures etc...

Stay with tested true RAID modes like RAID5 or 6, or even better RAID1 or 10

Also the Raptors are quite old drives nowdays, I can really recommend the Seagate Cheetah 15k.7 drives, they are quite affordable and really puts the Raptors to shame...

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

Seagate Cheetah has been around since SCSI-II and Raptor is old?

The Raptor line is still in production, so how is that old?

And $150 vs $250 [for the same storage capacity] is affordable?
http://www.newegg.com/Product/Produc...82E16822136322
http://www.newegg.com/Product/Produc...82E16822148538
.

[Uranium-235]

Re: Call to anyone with RAID 1E experience

raptor line is old and discontinued VELOCIraptor is the new offshoot line from it

the seagate cheetah line is still in development, and 15k sas drives can outperform velociraptors, which might be worth the extra $$$. but if you already have raptors, I would say just use those

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

raptor line is old and discontinued VELOCIraptor is the new offshoot line from it

Same line as far as I'm concerned.
Most other people too.
http://en.wikipedia.org/wiki/Western_Digital_Raptor
-
If that small a change [1.5G/s -> 3G/s] makes it a new line then when Cheetah went from U160 -> U320 [and every other transition] then it was a 'new line' each time.
.

[Uranium-235]

Re: Call to anyone with RAID 1E experience

true but compairing a velociraptor to a sas 15k drive is appes and oranges. cheetas are built to a higher standard. VR's are for enterprise and enthusiast use, you're more likely to find cheetas in server storage then vraptors

that isn't the only difference. vr's are 2.5", ice pack cooled and are generally faster, neither can really reach up to the sata limit (150 then 300)anyways

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

true but compairing a velociraptor to a sas 15k drive is appes and oranges.

Which is why I objected to the suggestion.

neither can really reach up to the sata limit (150 then 300)anyways

Cheetah's don't reach their interface speed either.
Head-disk transfer [media transfer rate] is the bottleneck.
- Which incidentally for a Cheetah 15k.7 SAS series is only 122-204 Mb/s. [Looked it up.]
-- Still not even close to SATA 1.5 Gb/s interface speed.

.

No, appes and oranges don't look alike at all.
.

[Digital Technophile]

Re: Call to anyone with RAID 1E experience

I don't have experience with that particular RAID mode but I would stay away from it because there are many properiatry modes like that, from different manufacturers
Like RAID 50 etc etc

What they all share in common is that if you get in trouble you really are gonna have a hard time to fix it, since it's such a special setup there wont be much info on the web about recovery procedures etc...

Stay with tested true RAID modes like RAID 5 or 6, or even better RAID 1 or 10

Also the Raptors are quite old drives nowdays, I can really recommend the Seagate Cheetah 15k.7 drives, they are quite affordable and really puts the Raptors to shame...

Thanks for reply Per. In creating this thread and all of the thoughtful replies with regard to RAID modes, I've been "back to school" scouring the 'net of any and all articles I can get. I just found this one, which explains RAID-1E better than IBM's site (they're the ones who came up with it in the first place)! And reviewing IBM's site, it occurred to me to see what hardware they implement this mode on. And to my surprise, IBM's ServeRAID family of adapters (SAS/SATA) are based wholly on LSI's host bus adapters! And in reading that byte pile article it has become apparent that RAID-1+0 is still superior to this new hybrid RAID-1E (especially if the 1E array contains an odd number of disks). So, I'm feeling pretty good about RAID-10 at this point, but I'm still in consideration of RAID 5 & 6. What is attractive about these two modes is their error correcting capability. For me, it is not enough that that an array is simply "redundant", I like the idea that even through the overhead, the data can remain safe from corruption (a far scarier scenario than simply losing a disk, IMHO). BUT...I feel with the proper hardware overhead can be minimised; and "proper" in the sense that a healthy amount of cash (> $1000) will need to be expended to obtain an IT strength card. As far as "proprietary" RAID technologies go,RAID 1E, 5+0, and 6+0 are becoming more mainstream in the high-end hba's. I would go as far to say as long as a card featured "ROC" (RAID On Chip), I would trust running these modes. That only leaves the last variable in the equation....

....the physical drives (truly the weak link in the chain). As I have no doubt discovered, name brand alone doesn't guarantee reliability. Here, a good measure of discrimination is called for. As my signature proudly states, mechanical hard drives will never be able to saturate a SATA III link period. Not 10k RPM, 15.7k, ever. SAS is a new one to me, I've never had real world experience with them. Yet I see there are SAS 6gbps drives offered?! I'd really like to see some empirical evidence of the drives sustaining that kind of load. As far as apes and oranges go, WD raptors, while not coming close to saturating a SATA II link, still have a measurable improvement over standard 7200 RPM drives. Cheetahs are definitely better suited for SAS (due to their history in the SCSI segment) anyway so there truly isn't a comparison to be made (though they have some bragging rights in the IOPS and random read/writes dept.). BUT as far as speed comparison: SAS = SATA II, and U320 SCSi a tad faster; but again show me a drive that can saturate these links. Haven't seen one yet and probably never will. This is the domain of Solid State Drives!
I can still find sealed 80GB WD800ADFS Raptors available for a nice low price, and if I'm really feeling adventurous, I might even scoop some 160GB USED WD1600ADFS drives I've seen there (hey that's what redundancy's for right?). It would be great to try some SAS drives for tickles and grins, but I'd rather direct the funds to a nicer HBA that support both SAS and SATA and be able to upgrade in the future when budget permits. And Velociraptors I scratch my head at trying to understand the reduced size of the platters. As a machinist there's an interesting effect of diameter to RPM ratio called surface speed per minute. Simply stated, a 3.5" platter (for sake of argument) spun at 10k RPM is moving faster at it's periphery (outside edge) than a 2.5" platter spinning at the same speed. I can understand the vibration and balance issues involved, but with the advent of fluid bearings, these effects are minimised considerably. But hey I'm no engineer, so I can't speak with authority in that realm. I will say though I still don't like the ultra high areal densities being bandied about by hard drive manufacturers. I don't care how good the engineering may be, there's just to much complexity and risk involved with owning terabyte HDDs. For me, any new drive I acquire will be <500GB.

Thank you all again for your insights and comments, keep 'em coming. It's refreshing to discuss these things with people interested in and knowledgeable in these subjects. And check out that Bytepile article, it's a very good read into those RAID modes.

[PCBONEZ]

Re: Call to anyone with RAID 1E experience

And Velociraptors I scratch my head at trying to understand the reduced size of the platters.

Power consumption.
They only use about 6 watts. [The Cheetahs are a bit over 16 watts.]
Sounds small but 10 watts x # of drives in an array adds up fast.

Consequently I scratch my head about the need for that heat-sink frame thing.
I think the REAL reason is simply to keep people from putting one in a laptop because Velociraptors use 12v and laptops don't provide that for drives.
.
.

[Digital Technophile]

Re: Call to anyone with RAID 1E experience

Power consumption.
They only use about 6 watts. [The Cheetahs are a bit over 16 watts.]
Consequently I scratch my head about the need for that heat-sink frame thing.
.

Yeah, it's definitely a lot less power to spin those tiny platters. But by the same token that should be producing less heat, thus negating the need for "ice pack" cooling mechanisms. I suppose that's a fancy way of saying 2.5 to 3.5 drive bay caddy

[Uranium-235]

Re: Call to anyone with RAID 1E experience

Which is why I objected to the suggestion.


Cheetah's don't reach their interface speed either.
Head-disk transfer [media transfer rate] is the bottleneck.
- Which incidentally for a Cheetah 15k.7 SAS series is only 122-204 Mb/s. [Looked it up.]
-- Still not even close to SATA 1.5 Gb/s interface speed.

.

No, appes and oranges don't look alike at all.
.

actually, for once, you're wrong, by 8 (factor of)

122-204 MB/s

one site has it average 166 MB/s = 1328 Mb/sec = 1.3Gb/s, almost to 1.5 gb, but it bursted to 200 MB, which is past 1.5

[severach]

Re: Call to anyone with RAID 1E experience

What is attractive about these two modes is their error correcting capability.

I had a flaky 120GB IDE hard drive that would corrupt data without the drive or controller noting any problem. I diagnosed it by testing copied ZIP files. When I got a SMD tool I replaced the RAM chip and the drive works fine.

I looked high and low for a disk RAID level that would use ECC to help ferret out failing drives that don't know they are failing. Raid 5 and 6 aren't it and none of the others are either. The industry seems unconcerned that drives can be something other than perfect or non working.

Raid 5 buys space at the expense of convenience and speed and Raid 6 uses extra drives to keep the speed up with failed drives. Parity was clever back when hard disks were more expensive than card processors. It's counter productive now that hard drives are so cheap. Once you give up on Raid 5 as a relic from the past a lot of el-cheapo RAID cards without hardware XOR become practical and you can spend the money you'd waste on a phat RAID card on more drives.

Raid 1 Mirror FTW!