Build a seperate system, run windows server or linux and run RAID1

found someone selling
P4 630 Prescott 3.0GHz 2MB L2 Cache with Asus P5LD2 motherboard
for $80bucks
Add some ram and i’ll be set.

or

Adaptec SATA RAID controller, model AAR-2410SA for $100

add it to my existing system

I have no clue how controllers work, but everyone raves about hardware raid.

So OT, whats my best choice?
Wait, why are you blaming the RAID? You said iTunes fried your music collection.

on a separate computer, why the hell not?

Hardware RAID for RAID1 isn’t going to be a huge difference. A little bit of performance, but not a great deal. Hardware RAID controllers help substantially when there is parity involved.

why?

RAID1 does not have parity So I’m not sure why you’re talking about the benefits of discrete raid controllers when parity is involved — because that isn’t remotely relevant to raid level 1.

Personally, if you KNOW you’ll never want more than 320gb, then just build a file server, and use a discrete card and a level 1 raid. BUT make SURE you use a decent DISCRETE controller. onboard == fail and aids, unless you’re talking higher-end server mobos.

However, if it were *ME*, I would get some extra drives, and run RAID5.
on a separate computer raid 1 is fine. on the primary computer with a raid card, you should always have a separate backup

I never said RAID1 had parity, I said hardware controllers help substantially when parity is involved, not that RAID1 had anything to do with it. He asked if a hardware raid controller would be beneficial to him, I have him an answer
okay well i think it would have been prudent to specify that you felt that raid5 and an extra hard drive would be benefitital to take advantage of that parity.
I don’t get your hate for onboard RAID 1 solutions – I’m not sure how it can fail so badly that you would have one working drive wors case scenario.

the drive failing isn’t the big downfall of onboard raid solutions — it’s the controller failing.

vhat?
parity = protection against one disk failing or am i wrong?
in raid 1 one data is mirrored against all the drives .. therefore any amount of drives can fail except 1, and you still have all your shit?

im wrong?

yes you’re wrong. what you describe is redundancy. parity is one method for acheiving that redundancy.

so parity implies redundance, but redundance does not imply parity.

ok, what is parity?

ensuring all the data is on at least 1 drive?

no

also, for a raid 1 (for my OSes) what’s wrong with software raid? is it fail and aids? or is it simply slower? by a lot?
how much would a cheap and reliable controller cost?
even if im already doing a raid 5 – I should have 2 raid cards?

i dont think i will setup a raid 1 because i will prob have 3×500gb drives in a raid 5 that ill fit my oses on, but in the future if i begin to find it tight for space, i might buy 2×100gb SSDs (this could be 5 years down the road) to relieve some space in my raid 5 and also do a lot to improve performace.
dont worry today about something you MIGHT do with your computer in 5 years. that’s just stupid.

im trying to understand raid, i asked a question about software raid

edit: still dont get how raid 5 does parity and raid 1 doesn’t. it seems pretty important so i dont think ill do a raid 1, but how does raid 5 have it and 1 doesnt?
edit: im slightly getting it
edit: yes.

raid 1:

A RAID 1 creates an exact copy (or mirror) of a set of data on two or more disks. This is useful when read performance or reliability are more important than data storage capacity. Such an array can only be as big as the smallest member disk. A classic RAID 1 mirrored pair contains two disks (see diagram), which increases reliability geometrically over a single disk. Since each member contains a complete copy of the data, and can be addressed independently, ordinary wear-and-tear reliability is raised by the power of the number of self-contained copies.

raid 5:

A RAID 5 uses block-level striping with parity data distributed across all member disks. RAID 5 has achieved popularity due to its low cost of redundancy. This can be seen by comparing the number of drives needed to achieve a given capacity. RAID 1 or RAID 0+1, which yield redundancy, give only s/2 storage capacity, where s is the sum of the capacities of n drives used. In RAID 5, the yield is s * (n – 1)/n. Using 1 TB drives as an example, four of them can build a 2 TB redundant array under RAID 1 or RAID 1+0, but they can be used to build a 3 TB array under RAID 5. Although RAID 5 is commonly implemented in a disk controller, some with hardware support for parity calculations (Hardware raid cards) and some using the main system processor (Motherboard based raid controllers), it can also be done at the operating system level using Windows "Dynamic Disks" or with mdadm in Linux. A minimum of 3 disks is required for a complete RAID 5 configuration. In some implementations a degraded RAID 5 disk set can be made (3 disk set of which only 2 are online).

In the example on the right, a read request for block "A1" would be serviced by disk 0. A simultaneous read request for block B1 would have to wait, but a read request for B2 could be serviced concurrently by disk 1.
so if there are parity errors reading, then, i guess the data can be retrieved off another disk and hopefully the parity matches, yes?. are errors reported and how?

it’s good for backup in terms of hard disk redundancy. if itunes fubared his music collection, it would do so on both raid disks. it doesn’t make historical backups.

if you’re making yourself a raid for home and you have 2 extra disks, raid-1 is fine. raid-5 requires a 3 hard disk minimum.

depends on the controller. one of the reasons a good controller > onboard.

parity is used for error checking, as well as to rebuild a data set if a drive fails. errors are generally reported, yes. most controllers have some way of accessing the data. often snmp can help collect data and it can be analyzed with cricket/mrtg or other statistics packages.

Yeah, but what’s the big deal if it does? You buy yourself a separate card and rebuild the mirror?
Okay people, get off the RAID thing. He could be running a server cluster connected to a RAID60 SAN and it wouldn’t matter, because what iTunes does to his music collection is its business; the disk, whatever kind it may be, is just going to do what it’s told to do.

But since it’s a going concern, I might as well throw in my two cents: RAID1 is fine for backups even if it doesn’t have parity, because instead it has a complete second copy of the data. You don’t need parity to reconstruct the data in case of disk failure, not when the entire data is redundant.

The problem is that if the onboard controller fails, what are the chances you’re going to find another controller that can detect the existing array and bring it online again without having to wipe the disks?

I, for one, use a RAID so I don’t have to make backups of non-important files, but if the controller fails and I can’t just swap in another identical controller and I end up having to wipe the disks and make a totally new array, then there goes my evil plan.

(I also use it for speed, but that’s not the issue here.)

But iTunes can still kill his music in the same way? = not a good backup plan.

Plan I’m liking is to get 3×500Gb drives in a RAID5. 1Tb storage = loads, to store my OSes, backups of my OSes (including laptop) updated weekly (+1 old to keep incase), my downloaded shit that I will probably find somebodies external harddrive to backup to, but backing that up isn’t worth worrying about, and my important stuff with 4 or 5 spares locally (will be just a few dozen MBs max), and as many as fills my quota at off-site storage (it isn’t my hard disk and I don’t need to listen. I might shop around and pay, or get off a friend, maybe do a deal) – this done nightly or on shutdown.

So I don’t have to fuck around too soon I might fit one hot spare. It doesn’t draw power and can’t piss me off, right?

Anyhow, OP do RAID 1 but if you wanna come back from another iTunes fuck up with your ratings etc (your iTunesDB I would think), you need to back it up. That doesn’t mean mirroring it, when stuff is deleted on a mirror it is deleted on all mirrors I would think (is this right dues?). I suggest you automate copying it to a seperate location on ur RAID 1 with your music, I’m sure there’s programs to do that.

Don’t see the need for a new system.
I agree that any kind of RAID is inappropriate for backing up important files; it’s just a good way to prevent data loss due to disk failure. Nothing can keep you from deleting something by accident, after all. That’s why I said he shouldn’t blame the RAID controller for losing his music in the first place.

Regarding the hot spare — it’s called a "hot spare" for a reason. Yes, the hot spare is on and spinning whenever the computer is on, unless you have some fancy RAID controller that’s smart enough to spin down the disks when it’s not using them.
all this confusing shit wont matter when we’ll switch to solid state drives in the next 2 years.

bullshit. RAID will be as important as ever.

That’s a bit shit. How fancy does it need to be to know that a configured hot spare isn’t needed until there’s a fail?

Even I know that!

lmao

you’re an idiot. it’s a hot spare because it can be kicked into service on-the-fly, and without user intervention.

hard drives CAN spindown when not used.

True enough, but a broken mirror can still be read by another controller – hell even a non-raid controller. You’d have to rebuild the mirror, but that’s a pretty moot point.

Not by themselves. They only respond to commands from the controller.

I’ve got a hotspare in my file server, and I just went into the server room and yanked it. When I tilted it side to side, I could clearly feel gyroscopic action from the spinning platters.

The RAID controller in my file server is a PERC6/i, so that should give you a benchmark. Those cards cost $1000 apiece.
Wow I thought my thread died, so never checked back.

I ended up getting the raid card. I know itunes can still fuck it up, I will keep another copy on an external periodically.

One question about raid 5, I have two matching 320gb right now. Can I add another 320gb from another brand? Cause I can’t find another of the same model that I have.
You can’t do RAID 5 with only two drives, so it’s a moot point.

Edit: Whoops, thought you were adding it later. Yes, you can add a diffferent model. In fact, it’s probably better that the drives are not from the same batch anyway.
It’s better to use the same model of hard drive, because they will respond similarly. 5Gen’s right about avoiding the same batch, though, because that will keep them from failing similarly as well, at least in theory.

Related posts:

1. My mobo died :wtc: One day when rebooting a windows update it all went...

i noobed myself by trying to let the mobo handle the raid.
mobo raid from most consumer mobos is fail and aids.

the only mobos that do decent raid are server-grade boards.
here’s my recommendation… assuming you mean SATA-II and you have a PCI-Ex4/8/16 slot available.

I have a rocketraid card and I love it. I would recommend them as well. I think P07 was the one who recommended it to me in the first place and I’m glad I got it.

I’d rather have an Adaptec/HP/Intel… but those are all 3x the cost…. For most people (including my personal workstation) the RocketRAID cards are an excellent value!

SATA-II
PCI-E

dunno about the other stuff.

would be nice to find one that is easy to expand and swap drives.

——
P07, how long would you guess it take to transfer a 4gb file from your desktop to the raid setup (on your workstation)?

.
thanks!

x8 and x16 slot compatible

how would you expand it out to 16?

it’s a x4 card…. so it will only use 4-lanes regardless of what slot you put it in. But a x4 card will physicall fit (and work) in a x8 and x16 slot — although there is no performance benefit for doing so.

kinda sorta. Think of a highway. Each car represents data. The more lanes you have, the more cars you can keep moving quickly, and the more throughput you have.

PCI-Express lanes provide approx 1.5Gbit/sec of usable bandwidth. As a result, a x4 card has a theoretical bandwidth of 6Gbit/sec or 768MByte/sec. That’s more than sufficient for a RAID controller.
Are PCI-Express lanes truly parallel, or can they combine?

parallel lanes essentially imply bonding.

Parallelism in general does not imply bonding. That’s why I asked.
i am not thinking of an instance where parallel would not imply bonding… I think you’d have to explicitly say that it didn’t if that were the case, as bonding is by far the norm.

but to directly answer the question: yes.
are you maybe confusing parallelism between discrete cards vs parallelism between lanes used to access a single card?

they are two different concepts.

Even with 8 3gb/sec drives, this is enough bandwith to not be a bottleneck? Could you explain this a little more or point to a good article?

Also, whys a discrete card > onboard RAID?

a SATA-II interface supports a maximum bandwidth of 300MByte/sec. The reality is that drives cannot actually obtain that speed.

Toms Hardware (I know, I hate them) found the FASTEST drive to be a WD VelociRaptor 10k RPM 16MB cache drive, and that could only get 102MByte/sec read performance.

The very highly recommended Seagate 7200.11 32MB cache drive (my personal favorite) only sustained 81.9 MByte/sec.

The popular Seagate 7200.11 8MB cache drives rated only 63.4 MByte/sec.

So lets assume you used the Seagate 7200.11 drives… 8 drives at 81.9 MByte/sec only use 655.2MByte/sec of bandwidth (max… assuming all drives are running at redline). So the RAID controller still has a nice margin on it.

onboard raid controllers use system ram. Many discrete controllers use their own, and some even allow you to insert a stick of memory in them. onboard raid solutions are almost exclusively the shittiest software-based implementations available. Even the "better" onboard raid solutions are shitty promise controllers. EMF, inductance, and refraction are common problems with onboard raid controllers, so performance is often much less than with a discrete controller. onboard raid controllers are also an extremely common failure point for mobos. replacing the mobo means a new controller, which means you generally lose your raid arrays. discrete cards make it easier to replace components without losing data — even if the controller goes out, you can often get a new one. you also get advanced features such as NCQ with many discrete cards — something that’s not generally available onboard (many mobos have ncq on non-raid channels, but not for raid.

i’m surprised to see you endorsing highpoint

but yeah they make some cool shit for a good price from the little experience i have had with their product. thats what i’d be looking at.

Are Raptors and VelociRaptors the same thing, or is the VelociRaptor a newer model?

velociraptors are new, been out for like a month or so IIRC
Hmm. I wonder how much better they are compared to my Raptors.

I’d like to see a 10krpm 2.5" SATA drive. I know they have them for SAS, but I’m not that rich.

EDIT: Oh shit, it is a 2.5" drive! Now allz I need is four of them plugged into this bad boy:

meh I’m not interested in 2.5" drives for desktops…. I like the 3.5" formfactor. It’s not too big, yet allows more capacity, and more importantly — better cooling over 2.5" drives.

According to that benchmark, the old 74GB 16M cache 10k RPM Raptors actually did WORSE than a basic 7200.10 Seagate

The 7200.10 did 79.8 MByte/sec
The Raptor did 75.3 MByte/sec

To add insult to injury, the Western Digital Caviar 750GB 16M cache drive did 75 MByte/sec….

It seems the VelociRaptor is legit, but the older Raptor was a marketing scam (which coincides with my personal results that showed a Raptor wasn’t worth the cost diff and lack of space).

Not really. The smaller something is, the better its area:volume ratio is. Heat can escape better from 2.5" drives, you just don’t notice in laptops because they have such shitty ventilation. The enclosure I posted has integrated cooling fans.

Not to mention, 2.5" drives have smaller platters and smaller motors, so they use less power and generate less heat in the first place, and the smaller platters also means the reader arm needs to travel less in a worst-case read scenario (i.e. arm on the inside circumference, data on the outside circumference, and v/v), which contributes at least a small improvement in random-seek speed.

So other than that, and the fact that enterprise systems are uniformly moving towards 2.5" disks as a result of those advantages, you’re completely right.

no. SURFACE AREA makes a huge difference. Making things smaller (such as processors) help because less power can be used to get data from a>b. But with hard drives, having a bigger surface area helps.

Take a processor and install a small heatsink…. runs hot.
Take that same processor and install a big heatsink…. runs cooler.

Enterprises are moving towards 2.5" disks for physical SIZE and POWER considerations. Notice they’re also moving towards blade datacenters. But they also run hotter than their 3.5" counterparts. That’s why cooling is also being increased.

That’s sustained-read speed. There’s not much improvement to be had over that, because you can’t cache reads and the bit-density of the platters has more effect on the throughput than anything else. I bought my Raptors to improve random I/O speeds, and my tests with those drives installed showed a marked improvement over the Seagate I had before — 3200 to 5700 PCMark points on the HDD subscore, despite the RAID controller sucking data through a PCI straw.

I’m sorry, but you really don’t know what you’re talking about. The area:volume ratio is paramount when transferring anything (be it heat from a heat source or sugar into a cell) through a barrier, which in this case is the surface of the hard drive.

Yes, if everything else is equal, if the drives make the exact same amount of waste heat and it’s all in direct contact with the outer casing of the drive, then the bigger drive will cool off faster due to more surface area. But everything is not equal; the smaller drive is cooler and the smaller size puts the heat sources closer to the casing, so the smaller drive will dissipate heat better.

You lose. Try to deal with it.
random I/O speeds has more to do with seek time than anything else. The 10k rpm is what gives the raptor it’s advantage in that. Go with a 15k SAS drive and it’s even better.

I don’t see why Seagate doesn’t release a 15k version of it’s high-capacity drives.

Too much heat will increase the platters’ suceptibility to superparamagnetic corruption, and then all that data is gone.

Well I just went to Seagate.com and they list the operating temperature to be 5 degrees celcius HIGHER with the 2.5" version of their 7200.3 2.5" drive (current model) compared to the 3.5" ES2 "enterprise" model.

Same company. Same capacity. Same speed. both recommended for enterprise workstation/servers…. The smaller drive runs hotter.

can’t be any more than a velociraptor…. i mean what I want seagate to do is the same concept.

Then they’ve done something to improve the heat conductivity of the larger drive, perhaps by equipping the larger drive with a single platter instead of two, or something like that. Physics does allow for that, but it’s just a loophole in the rule that if you have two black boxes, one larger and one smaller, with equal heat sources at their centers, the smaller one will stabilize at a lower temperature because the heat can escape more easily.

The Velociraptor will make less heat, though, because its motor is smaller. To do the same with a larger drive equipped with a larger motor, you’d have to decrease the bit density to prevent heat-related failure.

seagate has 2.5" drives…. WD expanded their 2.5" offering to something that was too tall to fit in a notebook, and enclosed it in an aluminum heatsink to bring it to standard 3.5" desktop drive specs. I don’t see why Seagate couldn’t do the same.
Its really annoying that ZFS is there and available so this guy doesn’t need a RAID card – but Sun keeps it locked up by making it GPL incompatible. Then media could be hosted on a linux storage appliance and he could access it however.
i fail to see how ZFS would solve his problem. care to elaborate?

RAID solves three issues:
1) Redundancy — drives can fail, and data is not lost. In fact system keeps going while you replace the drive.

2) Capacity — you can fit much more data on a volume than is typical from a single drive.

3) Performance — you can read and write data faster.
ZFS solves all three of those issues for a home setup. Look up ZFS.

I know the basics of ZFS. It has many cool characteristics. My issues, however, revolve on it’s "RAID-Z" being software-based… Which signifigantly reduces it’s usefulness for various operating systems. There are also performance concerns that I would have — although I don’t know where ZFS ranks on that. Also, my understanding is that there is no quota system…. lame. Also, I don’t think you can add a new disk and grow a RAID-Z implementation.
ZFS doesn’t have the problems that traditional software raid does. And yes, you can grow a raid-z array.

i recall it being a problem that you couldn’t add drives once it was created.
Worth it to take a look on eBay for Perc5/i pulls. Those are just badged LSI MegaRaid 8408s. Pretty easy to find with BBU and cache for $120 or so.

SATA?

Yes, SAS/SATA

I have 2

The ZFS Boot project successfully added boot support to the OpenSolaris project in March 2007.

But its not in Solaris yet, I don’t believe. You would expect ZFS using appliances to run OpenSolaris. Or BSD. Or even OS X when that port hits. Probably the mac mini will allow you to have a zpool for your media storage eventually.

But ogre is right: you can’t dynamically grow a raidz zpool just yet. You can with JBOD/mirrored ones, but not striped.

SATA and SAS. You can get 15K SAS drives for $200 bucks, 3 of those in a RAID-5 screams, I sustain 200MB/s R/W.

why don’t i see any ports on the card(s)?
There should be two ports in the rear. You need 1 or 2 breakout cables to plug in your drives, 4 drives per cable.

EDIT: Looking at the ones I see on eBay, they either have external ports or on the top behind the memory.

So you know, breakout cables can be pretty expensive.

Adaptec has some out for <$20 for 1m. They’re pretty nice.

I’m running a Perc5/i too.

Damn. I know that when I got my cards, the one I couldn’t use had one 2 breakout cables with it. The other 2 I have didn’t come with any cables and when I was looking online I was seeing ~$100 per for cheap ones. I ended up just getting some nice but long as hell ones from the place I got my cards at for free.
With the perc5, I will have to run the breakout cables from the back of my tower (the ports only appear to be on the outside), back into the tower to connect to my drives?

Not necessarily. They have ones with external and others with internal connections. Just get an internal one. I’m not positive but looking at eBay it seems the Perc5/e is the external and the Perc5/i is the internal ones (gotta love the logical naming). Just check the descriptions, it should say.
The Perc5/i has no external ports. The Perc5/e has external ports only. The Perc5/e is usually significantly more expensive than the 5/i too.
so why do you need a raid card? im running my 1st raid0 setup with my Asrock motherboard and its worked ok say far……besides if its raid0 couldn’t it bottleneck at the PCI ports? running multiple drives?
running raid5 with four 500gb WD drives using the raid setup on my motherboard. want a nice card so i can expand the raid when needed.

shit is slow to say the least.

Because onboard RAID uses the CPU instead of bringing its own processor to the party, and because if it dies, good luck finding another controller that’s compatible.

With crazy numbers of cores happening – how long before RAID cards are obsolete? Why NOT do RAID calcs in one core, or several, if we’re going to have so many? Yet another reason why ZFS is so darn neat.

There’s a couple of reasons why general-purpose hardware will never replace dedicated hardware:

1. General-purpose hardware is less secure. I don’t want my RAID controller to jump from core to core as the operating system sees fit, I want it to keep its head down and do its thing, undisturbed by anything else going on in the system. You can just imagine how much fun hackers would have trying to write viruses that fuck with ZFS RAIDs, or RAIDs running Microsoft’s inevitable proprietary clone of ZFS.

2. General-purpose hardware is slower. Yes, CPUs are getting faster, but why would I want to give up any of my processing power to what should be a transparent background process, and why would I want to run that transparent background process on something that gives me lower throughput than a dedicated card would have?

3. General-purpose hardware is less stable. Remember WinModems? Yeah.

4. You can’t boot from a software RAID; the array needs to be a black box as far as the OS is concerned, otherwise there will have to be a non-redundant partition that contains the RAID drivers. I’m not interested in such nonsense.
This sounds exactly like an argument for a mini-computer. And 4 is just plain wrong.
thanks for all the help.

i ordered the Perc5/i, cable, and scored three 1tb drives.
I am thinking about grabbing a Perc5/i tomorrow.

That type of cache/ram does it use?

Does Vista include needed drivers for the card?

And is there different types of Perc5/i or are they all the same?

Also what type of cables does it use?

Correction: The only way you can boot from a software RAID is if the boot loader, the kernel, and the RAID driver can fit in a single stripe on a single disk. Which is to say, it’s a hacked solution.

As evidenced by the built-in hardware microcontrollers that ALL hard drives use nowadays, it’s better for storage management to be delegated to something that doesn’t have anything else weighing on its tiny silicon mind.
Yeah. And Math should be done on a math co-processor. Its OBVIOUS, isn’t it?
You guys take too long to reply, so i ordered it anyways

DDR…forget what speed

Doubt it

There are two types. Perc 5/i and Perc 5/i Integrated. The vast majority on ebay are the integrated type. Functionally they are the same.

SFF-8484 -> whatever your backplane or drives use. Generally it will be SFF-8484 -> 4x SATA fanout cable. Adaptec makes a couple of those cables that are pretty nice for quite cheap. I have an extra one I’ll sell you.

I believe it uses DDR2 according to the other sources I found…

I bought this cable, will it work?

Where would I find drivers for the card?

You’re right it does. I just remembered and was about to edit my post.

Yes that is the type of cable you want.

Dell has them. They’re listed under PowerEdge 2950 if you need Win 2k3 drivers or Precision 490 if you need XP or Vista drivers.

It is done on a math coprocessor. Math coprocessors used to be separate, but even though they’re on the same die as the CPU nowadays, they still consist of dedicated circuits that do nothing but whatever the hell math coprocessors do.

If you design me a CPU with an integrated storage processor and 64MB of dedicated RAM, I’ll consider using it.

Sweet. Thanks.

I don’t have to design it. You’ll be getting the functional equivalent in a few years from Intel. And ZFS or a child of ZFS is the filesystem you’ll be running.
No, it won’t be the filesystem I’m running. Because I won’t be running a filesystem that will be redundant with my dedicated RAID controller, regardless of whether it’s on the CPU die or in a PCI-Ex slot.

i paid around $30 for the same cable!

correct 490 driver?

or

The first one is a Windows Server 2k3 driver. The second is a firmware update for the card.

Dell’s support site just started acting up for me or I’d give you the R numbered filename to look for.
if install a new OS will all the data on the raid be safe?

the card was real easy to install and setup!
The data on the RAID will be as safe as the data on a single drive. The only thing you need to worry about is whether the new OS will have a driver to talk to the RAID controller. Even if you don’t have a driver, though, that won’t cause the RAID to fail, it’ll just make it inaccessible until the driver is installed.

I have an old Raptor and did some stress testing. A lot of poor mechanics that made the drive fail back and the day. Not to mention big heat issues.
It’s definitely a hot little bugger, but no hotter than comparable 10krpm SCSI drives. The biggest problem it has is that it was sold to people who didn’t realize that a high-performance disk would need actual cooling, and so a lot of the poor things just overheated and died.

I was one of them, to be honest. Fortunately, I wised up when I saw my case temperature readout, and I bought a hotswap rack with a dedicated exhaust fan. They’re still hot, but they’re at least stable now.

Back to this fight.

I started looking up software RAID on Linux and I found a writeup on the matter of software raid vs hardware from a linux engineer that makes good sense. There is no one solution to rule them all, but software raid seems appealing enough for me to seriously consider.

I might try it on this comp to see about the performance impact, I mean, I don’t have to spend. My trouble is however that my disks have irregular sizes, and I need to sort the bastards out aswell.
On a side note. I wonder if FDE (full data encryption) drives will come to the consumer market in the next few years.

I read the link. Thing is – ZFS isn’t software RAID. Its an evolution of raid and filesystems. Its a big leap forward. Thats why duex doesn’t get it, and its why we’ll all be using it or one of its descendents in the future. Read up on it. Its really neat.

"Equipped with a hi-speed 9500RPM double roll ball-bearing cooling fan"

wtf steroid fan?

9500 RPM is no big deal for a 40mm fan
Yeah, it’s not going to make much noise; it’s too small.

Size of the fan doesn’t matter when it comes to motor noise. I don’t give a shit how big a fan is, but a 9500 is going to be loud as fuck. Just from google, the NMB and Delta ones I’ve found are around ~42dBa, which is pretty damned loud.

Check out Sunon’s fans. They’re surprisingly quiet.

In any event, you can always put a slower fan in there if it bothers you too much.

there are plenty of SW /linux/open source based units and solutions out there.. FreeNas, Drobo, etc,etc..

My FreeNas unit is kicking ass for over a year on R5 and 8 drives now..

HMM you are right enough, I just read this about RAID-Z (VERY NICE):

I’m gonna setup a software raid, tidying my disks now. I feel pissed off about still the small possibility of corrupting much of my data explained in that blog. Will have to see if there’s anything I can do.

Although I do agree that it’s not the typical software raid logic, it’s not evolutionary by any means. Been done before on tru64 with advfs which is now open source, fs’ on vms, and wafl on netapp NAS’ (which is why there is a lawsuit).

i’m still not satisfied with it yet. i’m still telling everyone to stick to veritas, lvm2, and disk suite here.

All I want is my data to be safe, so what would you recommend to me?

I didn’t know lvm2 could do anything for me..

safe from what

Safe from disk failures, power outages, disk corruption and the like.

disk failures: raid 1, hardware or software

power outages: ups that tells hosts to perform a graceful shutdown after x minutes

disk corruption: chronic backups to another media, offsite if it’s important.

and the like: good admin habits.

They’re already here. Check out Seagate and Fujitsu.

If it’s the filesystem to end all filesystems, then I’m still going to wait for a native hardware implementation. I see no reason to waste CPU cycles doing something that every computer is guaranteed to do, when I could be using those CPU cycles running programs that only I care about.

EMC DMX + UPS + offsite = maybeeeeeeeee

no such thing in life as "guarantee" except death… (no taxes here)

There are many things every computer to a high degree of accuracy does and you’d be waiting a long time for hardware implementations. e.g. in the realm of the internet: html, javascript, xml decoding.

I prefer the idea of software raid because 1. no spending on raid card (Money spent on a raid card can go into memory and cpu upgrades), 2. disks don’t need to be equal size – as the OS can work on the partition level, incredibly useful. ZFS can work on it’s filesystem level for extra-cool data recovery (explained in that blog post).

I’m just unsure about performance impact and hardware bottlenecks.

thank you crontab. I might invest in a UPS.

Related posts:

1. My mobo died :wtc: One day when rebooting a windows update it all went...