Notebooks w/ RAID?

macemoneta asks: "Are there any notebooks available on the market that support (bootable) RAID (at least two 40GB+ drives as RAID0 and or RAID1)? While the rest of the components in 'desktop replacement' notebooks are quickly getting up to snuff, the hard drives are anemic in performance, capacity and reliability compared to desktops. Being able to use software RAID to create high performance meta devices and high reliability meta devices would really kick notebooks into high gear. Before anyone complains about size, weight, power and heat remember that notebooks have gone from 12 inch screens to 16 inch screens and 486 to P4M in the last few years. Most desktop replacement laptops use the batteries as a UPS, since they usually only last 90 minutes or less anyway."

Comments

Multiple drives?

[Linux_ho]

You would do well to simplify the question by asking if there are notebooks available with multiple hard drives. Lilo can boot to a software raid, so any notebook with two drives can do this. Whether the sound card or built-in ethernet/modem works is a different story, but you didn't mention whether those were important to you.

Re:Multiple drives?

[macemoneta]

I wanted to eliminate non-bootable RAID configurations: drives off a PCMCIA controller or Firewire drives, unless supported in the BIOS. I'm not sure that all the notebooks that support multiple drives can actually boot from any/all of them.

laptop drive limits

[OpenMind(tm)]

I'm not sure this is a good idea. Laptop drives, even the recent IBM enhanced models, are rated for a much lighter activity cycle than desktop drives. That is, push them as hard and as long as full-size drives , they are likely to fail on you. IBM is trying to fix this to make their 2.5 in drives suitable to blade servers. Still, RAID historically pushes drives hard enough to decrease the time between failures quite a bit. Combine this with drives designed for low load, and you're asking for trouble. I think the recommendation of a firewire external drive was a good one.

Re:laptop drive limits

[thejson]

I don't believe there's any reason to assume that using RAID would push the individual drives any harder than a single disk would be. I think this could be a good thing, especially for mobile desktop users. With RAID in a laptop, there would most likely be at most two disks due to space concerns. This would limit you to RAID 0 or 1. Consider RAID 0/striping: any reading or writing to the drive is split accross both drives, so in theory they should be doing half the work. However if one disk does fail, you're screwed, which is why this isn't a good option for a laptop (technically not even really RAID) Now consider RAID 1/mirroring. Each drive does the same amount of work as one individual drive would. There may be some additional overhead, but the added reliability is well worth it. If one drive fails, which is more likely to happen in a notebook, the problematic hard drive can be replaced without data loss.

Re:laptop drive limits

[OpenMind(tm)]

I'm not entirely sure of the full rationale, only that I've come across quite a few references to increased disk loading in raid systems, in addition to plain old working experience. I think the particular stripesize/cylinder size ratio, stripe layout, etc. have a lot to do with how much effect this has. A striped drive tends to end up doing a lot more short reads as compared to long sequential reads. Depending on how well the drive heads are managed (easier to do well in hardware, IMHO) the increased seek activity can generate a lot of wear and tear. This is not as big a deal when using the RAID with large files and a small set of data consumers. Again, this is my non-hardware expert interpretation, but there is a lot of reference in the literature to the shortened life of RAID drives.

Re:laptop drive limits

[macemoneta]

Using software RAID (in Linux, for example), you can do both RAID0 and RAID1 on the same pair of drives, for different (matching sets of) partitions.

You could setup a "work" partition set to be RAID0 for fast processing of large files (video data conversion, for example), while the rest of the partitions are RAID1 for protection from a failure.

With software RAID, it's not an either/or situation.

Re:laptop drive limits

[Shanep]

I've come across quite a few references to increased disk loading in raid systems

Out of interest, could you point them out?

I would have thought that disks in RAID systems are usually under increased load because RAID is often chosen for high load scenarios.

Kinda like saying red cars are involved in 20% more accidents than other coloured cars, therefore you have a 20% increased chance of having an accident if you drive a red car. Without considering that there are actually 20% more red cars on the road than any other colour.

PS, I just pulled the 20% figure out of the air to illustrate my point, I don't know what the real number is, just that there are more red cars on the road which can lead to ridiculously misinterpreted statistics and thoughts of voodoo and such.

Re:laptop drive limits

[OpenMind(tm)]

I've looked a little closer into the references I could find on the web about decreased MTBF on drives in a RAID scenario. It appears that the way most people are viewing this is as as statistical phenomenon. If you have 6 drives with a fixed probability of failure, the whole set has a higher probability than the individual drives. I could not find any references, in the last 15 minutes, to increased loading, so perhaps that part was a phantom of memory. Either way, with low reliability drives, dividing the MTBF by two is not the best idea.

Re:laptop drive limits

[Tipsy McStagger]

Yeah, but with a system thats back home infrequently might it be better to have a lower MTBF and the ability to repair the damage than a system that rarely breaks but when it does its catastrophic.

Re:laptop drive limits

[Shanep]

might it be better to have a lower MTBF and the ability to repair the damage than a system that rarely breaks but when it does its catastrophic.

Exactly.

If you don't need super performance, can't tolerate loss and can afford it, just mirror across some good drives.

SCSI drives seem to have higher MTBF's and longer warrantees, so after a bit of research amongst them you should be able to find some reliable drives.

If you need performance and cannot tolerate loss, RAID5 or RAID0+1 could be good for you.

But regardless of your disk system choices, choose a backup method that has a reliable restore feature. ; )

Faster, not more...

[TibbonZero]

I would rather have a 15000 RPM harddrive, than 2 4800 rpm drives in a laptop. Space, battery, etc..

Re:Faster, not more...

[autocracy]

No, you wouldn't. Go find a 15000 RPM drive. Plug in just the power cable, and give it some juice. Once it's spun up, unplug it, and then start moving it around. I'll tell you, it's very interesting to have a hard drive fighting you. The same effect would happen when you started moving your laptop. And of course, giving that thing a hit when it's spinning would ruin your day.

Actually...

[cmowire]

At my place of employ, we've been pleasantly surprised with the performance of the latest latest laptop hard drives, actually. Compares quite favorably to anything but 7200 RPM drives, at least for what we do (software development).

Depending on how you configure it, some Thinkpads can take two IDE drives. But I suspect your best bet is to get a pair of 3.5" drives in an external case and either hook up via FireWire or SCSI. I doubt that 2.5" IDE drives will take well to RAID 0.

Look at the Eurocom 8880

[compwizrd]

The Eurocom 8880 has the capability for FOUR hard drives at once

http://www.eurocom.ca/products/showroom/specs888 .c fm

Has the capability for two cdroms/dvd-roms/etc at once.

15.7" screen as well.

No mention of weight, I suspect you don't wanna know.

Eurocom has always been a little bit ahead of everyone else on getting things out :)

I believe the TV tuner option replaces one of the media bays though

Not the unix way

[bluGill]

Sorry, but I gotta complain about your definition of desktop replacement. I want a desktop replacement laptop because I can take it to class and meetings. I love taking notes on my laptop for 8 hours without plugging in. Meeting minutes should be sent out within 1 minute of the meeting ending. When someone is in a meeting they should be able to get at the unexpected data they need that is on their desk, without leaving the room (that is the data is on the computer, so any computer can access it), not run back to their desk to price off a hard copy. Sure the batteries are a UPS, but if that is all you use your laptop for, a desktop with a UPS is cheaper and has a better keyboard, and monitor.

When I need more power than is on my laptop, then I ssh over to one of our CPU servers (at my last job we had a couple machines set up for this), and display my apps back to the laptop.

Laptop users should not have a big harddrive. They should have enough room for the OS, a few apps that they run on the road, and a copy of the documents they use often. Note I said copy, the master copy of these documents should be someplace that is always backed up, the laptop just has a copy for quick work until it can be synced with the master (two way sync). Other than that, long battery life, weight (the mass of even a heavey laptop is not enough to make a difference in gravity humans can stand, so weight is the critical), screen, keyboard, and interfaces are key.

Remember laptops are stolen; droped and broke; and forgotten far too often to have the master copy of anything. They should be easially replaceable. Too expensive to be disposable, but too fragile to depend on any one.

When you lock yourself into the desk is where I work, you miss the power that a laptop gives you. Get a laptop that you can use anywhere, wireless networking in the office (warning, security issues need to be addressed here), with a battery that is worth something. Suddenly the laptop goes from an expensive toy that to impress people to a useful tool that does things you couldn't get done otherwise.

Re:Not the unix way

[nomel]

Do you realy get 8 hours? What Kind of laptop you have? Cause that's real nice...

Re:Not the unix way

[bluGill]

Yep, a IBM thinkpad 390. It got just over 4 hours from a battery, and I had two. I only used the cdrom drive once in the year I had it, so removing the CD/floppy for a second battery was a good trade off to me. I was also careful to avoid using anything that took power, all the computing was done on the backroom cpu servers, and this was just a display.

Cheaper option

[bill_mcgonigle]

Wow, so many links to expensive hardware. Pick up a used Pismo powerbook (last black one). You can take out the CD drive, and replace it with another laptop drive. Sleds are available from VST, probably find some used ones too. OSX has software RAID-1 built in.

You could use any of the black line, but the Pismos often had 500MHz G3's.

4.9kg/10.8lbs

[Chris Pimlott]

Ouch. Weight is listed in the full specifications. Only 2 hours of battery life too. That's not too bad for all that's in it, though, and CDN$2999 doesn't seems too unreasonable a price.

SCSI

[Rheingold]

For a modern multi-tasking operating system like Linux or BSD, SCSI is a better answer than RAID. I've always found interactive use much better than with IDE, even than the newer ATA-100 and ATA-133 drives. IBM used to make RS/6000s with 2.5" SCSI drives and titanium cases. There were really sweet!

Huh?

[bellings]

Being able to use software RAID to create high performance meta devices and high reliability meta devices would really kick notebooks into high gear.

Can someone explain to me what a "meta device" is? And, can some explain to me what happens when you get a plurality of "meta devices?" I mean, what the hell is this guy blathering about?

I'm sure other people are going to point out that RAID probably isn't the solution to the problem you think you're having. It's like hoping that forged connecting rods and pistons are going to make your Dodge Neon into a sports car. By itself, it's not going to do what you seem to hope it will do.

Anyhow, if you want a good, decent, fast notebook get yourself and Apple.

Re:Huh?

[macemoneta]

No matter how fast a single drive is, striped drives are faster...

Re:Huh?

[macemoneta]

If you check the author of the question, the clueless one would be me. :-)

I used the term meta device, because that's how Linux uses the term for it's software RAID, which is what I'm interested in using. Physical devices are combined into meta devices (/dev/md0, /dev/md1, etc.).

In RAID0 both drives are seeking, but for different data. There's no requirement that the drives be synchronized in any way. Once the first buffer arrives to satisfy the request, the the application is dispatchable. Synchronized spindles were a technology that was used at one point, but the last I heard about it was about 15 years ago. FTR (full track read) in the drive firmware and similar competing optimizations eliminated the need for synchronization and the associated RPS (rotational position sensing), and RPS-miss condition you're describing.

In RAID1, either device can satisfy a read, so read processing is slightly faster. Both devices have to acknowledge a write, which does reduce performance slightly, for the increased reliability.

Re:Huh?

[macemoneta]

I've seen both terms used... Search groups.google.com for:

"meta device" linux or
metadevice linux

The source says "multiple device", so you're right; I'll use the correct terminology going forward. Thanks for pointing that out!

Side point

[macdaddy]

I bought an Asus A7V333 with the onboard RAID. Once per minute regular as clock work the load spikes to 30% on top of what's already running. Once a minutes, regular as clockwork. Up and down real quick. Games stall, MPEGs stall, etc... The problem is apparently the RAID. I have 4 120GB WD 8MB cache models. 2 as regular drives. 2 as a stripe on the OEM Promise FastTrack133 builtin controller. If I disable RAID with the jumper, no spike. If I re-enable, regular spike. If I leave enabled and disconnect the 2 striped drives, no spike. It's got to be the RAID. Asus won't return my calls. I guess this means I'll never buy an Asus again and probably never buy a board with onboard RAID again. A buddy of mine blames it on the Via chipsets. Could be.

That said, I'm not so sure you want to buy anything with onboard RAID. Perhaps you should look at a speedy Firewire drive.

Re:Side point

[macdaddy]

That's what I'm afraid of. I guess I'll buy another board. Perhaps this board will work ok as a Linux box. I won't be using all the fancy features in a Linux server. Now I just have to figure out what board to buy. Decent overclocking would be nice. Decent onboard features would be nice too. No Via anywhere please.

Re:What is your reason for wanting RAID?

[macemoneta]

I use software RAID on Linux; the reason I'm looking for it on a notebook. Performance is excellent and CPU load is negligible (even on an old PPro200) even during intensive I/O bound operations. You can mix RAID0 for performance/capacity and RAID1 for reliability on the same pair of drives (different partitions), providing exceptional flexibility.

You really should try it.