Page 1 of 1

Brainstorm with Me About Storage and Computing Solutions

Posted: Mon Nov 06, 2006 3:15 pm
by cycleback
I would like some help brainstorming for some possible solutions to my computing needs. I have about 900 GB spread over 9 IDE hard drives which are all essentially full. My case is also at its limit in the number of drives that can be stuffed into it. Some of the files on the drives are duplicated or burned to DVD but the vast majority are not and it is a disaster waiting to happen. The price of hard disk storage space is approaching that of DVDs especially if I factor in their ease of use and account for my time. I would like your help in coming up with ideas of how I could create redundant relatively safe storage on a cheap budget (I am a grad student). What ever I end up using I would like it to be relatively expandable and quiet since I live in a small apartment.

Overall Needs Ranked in Order of Importance (Further Items Might Not Happen Any Time Soon):
1) Storage Solution:
  • ~1 to 1.5 TB of redundant hard drive storage
  • Relatively cheap beyond the costs of the hard drives themselves
  • Expandable
  • Quiet
  • Easy managment (Preferably would appear as a single volume)
  • Low Power
  • Directly accessible from Windows (Samba share would probably work though I would prefer it to be directly connected)
2) Quiet low power torrent and download box ( I live in a small apartment with my desk about 3 feet from my bed)
3) Another machine to mount another BenQ 1655 so that as I am burning DVDs I can be doing PI/PO scans using Nero CDSpeed in parallel
4) Improved interfaces:
  • USB 2.0 - Neither of my computers has USB 2.0 I would like to get an IPod Nano soon so USB 2.0 connections would be useful.
  • Sata II - I would prefer not to buy anymore IDE drives as they are a dying technology and the cabling restricts airflow so much.
5) New LCD monitor currently I am using a dying 17'' CRT
6) New machine all to run my numerically and disk intensive simulations at home (FEA).


I have the following hardware currently available:

Desktop (Windows XP SP2) (This is the main machine that I use at home day to day):

AMD Athlon 1.33 GHz (CNPS7000B-AlCU CPU Cooler)
512 MB DDR 266 RAM
IWill KA266+ (USB ports do not function I think because I accidently cut some traces on the MB) (2 PCI slots left)
ANTEC SLK3000B Case (Yate Loon 120 mm fan mounted to cool 5 hard drives)
ANTEC Truepower 2.0 430 W PSU
Matrox Millenium G550
3Com Etherlink XL 10/100 PCI Ethernet Card
Promise Ultra100 Tx2 IDE Controller Card
Promise Ultra133 Tx2 IDE Controller Card
BenQ 1655 DVD Burner
NEC 3540 DVD Burner
Kingwin Mobile Rack
CoolerMaster 4 in 3 Hard Drive Mount (This is mounted in the top three 5.25 bays)

Laptop (Windows XP SP2):
Dell 4100 (600 Mz P3)
40 GB Samsung Laptop Hard Drive

Network Router:
Linksys WRT54G v4

I am leaning to some sort of RAID6 solution because of problems RAID5 has with rebuilds and uncorrectable read errors. I relize that RAID is not a backup but I have not seen a viable backup solution for ~1 TB at a cost I am willing to pay. As far as I can tell only recent PCIe or PCI-X hardware raid controllers support RAID 6 expansion. Except for the problems with growing the array the md linux solution looks to be the most cost effective unfortunately I would like something that I can expand to account for future storage needs.

A) Are there any other solutions that I should consider or gotchas that I haven't listed below.
B) Which way would you go and why?
C) If I go the Linux route and pick up an older motherboard and processor what would you recommend? What current heatsinks and fans would work with it? Is there a specific used components that I could ebay that would be relatively quiet?

Ideas that I am considering for storage solutions:

Solaris x86 using ZFS
Pros:
  • ZFS management sees relatively easy and well documented
  • Checksumming would help data corruption problems
  • RAIDz2 is essentially equivalent to RAID 6
Cons:
  • Limited hardware compatibility (It is possible I would need a new MB, CPU, and RAM)
  • Without a 64 Bit CPU the amount of memory that ZFS can use is limited to 512 MB and apparently has some impact on performance.
  • It is not clear if Samba works correctly? (It seems to be built in but there are problems with ACLs)
  • It is not currently possible to add devices to a raidz vdev. I couldn't expand the vdev by adding the same size disks.
  • Solaris seems to be lacking a good package managment system. Nexata looks interesting but it is still in beta and doesn't update the core OS.
  • I am not familiar with Solaris administration and updating.

Linux Software RAID5/6 on an Old Machine

I could pick up a cheap P3 motherboard, CPU, RAM, Case, and PSU, and some PCI sata cards and make a linux software raid with md and maybe EVMS.
Pros:
  • Potenially cheap hardware except for sata cards, case, and psu.
Cons:
  • RAID 6 arrays cannot be grown currently
  • I have not used Linux/Unix in a serious way for at least 8 years
  • No vendor support for RAID solutions
  • I have enough old hardware already
  • Unclear what type of bit scrubbing md does to avoid uncorrectable read errors
Infrant ReadyNAS NV+
Pros:
  • Easy setup and administration
  • X-RAID allows expansion as I can afford disks
  • Low power
Cons:
  • Limited to 4 disks (I wish this had 8 bays)
  • Cost (~$650 for the device without disks)
  • Less flexible than a general linux installation
  • RAID 6 is not available
  • Unclear what type of bit scrubbing the device does to avoid uncorrectable read errors
Hardware RAID Card:

Two recent hardware RAID cards support RAID 6:

Areca ARC-1220 PCI-Express x8 Sata II Controller Card - $494.00
3ware 9590SE-8ML PCI Express x4 SATA II - $504.00

Pros:
  • I could keep my storage directly accessible in a Windows machine
  • Easy administration
Cons:
  • Hardware compatibility lists for motherboards and hard drives
  • Battery backups add at least another $100 dollars
  • Cost and I would need to purchase a new MB, CPU, and RAM.
  • The hardware controller becomes a single point of failure.
  • Not as easily sharable to other computers in the future.
eSata Card + External Multiple Drive Bays use Synchronization programs to create redundancy between drives
Pros:
  • Relatively inexpensive
  • Potenially directly attached to my current machine
Cons:
  • Loss of half my storage space to mirror drives.
  • PCI bus in my current machine is already fairly crowded.
A) Are there any other solutions that I should consider or gotchas that I haven't listed above.
B) Which way would you go and why?
C) If I go the Linux route and pick up an older motherboard and processor what would you recommend? What current heatsinks and fans would work with it? Is there a specific used components that I could ebay that would be relatively quiet?

Posted: Mon Nov 06, 2006 10:12 pm
by Poodle
How much truth is really behind all this RAID-5 and SATA panic?

Posted: Tue Nov 07, 2006 2:17 am
by jaganath
You might get some better responses by posting this in the Storage forum? I don't even know what RAID is 8)

Posted: Tue Nov 07, 2006 5:23 am
by jjr
Tricky question ... I've had a similar problem myslef and came to some similar conclusion : if you factor the ease of usage etc. HD storage IS cheaper than DVD. Possibly less bad suprises as well as a burnt DVD can become unredable without you being aware. While a disk you'll now right away.

The only difference is I do not have too much of a budget problem and solved it pretty easily by stuffing a couple of 200MB+ HD in my old box.
(About 450Mb of mirrored space currently)

The way out is maybe to try to simplify the problem by finding some uncontournable facts to start with :
- if you need redundancy and current storage is full you will need to get more stotage space : the rest (which software and supporting hardware) won't change that in any way
- the most trouble free solution is possibly raid as you'll have no sync need of any sort,
- if you are not used to it, advanced forms of raid can become quite a pain at the time of recuperating the data (though definitely feasible from a technical point of view)
- you're data is safest if your system is off so dissociation the server from you bittorrent download is recommendable (I do that myself : a fileserver with all the storage space and occasionnal power up to back data from my laptop)
- any CPU above 800Mhz is more than enough for a small fileserve like yours (if you don't want to use your 1.3 Ghz PC you can use a wide range of oldies)
- a small box with many drives will be cheaper than any storage server solution (lan disc etc ...)
- you are close to the limit of the number of drives you can support so to support more you will need to invest in controllers : if you have old drives of limited capacity (can you provide a list ?) it is is probably cheaper to throw them away or use some extractible racks than invest in more controllers

Now some other hints :
- Raid 1 is the safest way but the one that will need more storage space,
- other forms of raid will save you SOME disk space, but in comparison with the trouble of handling it, it may not be worth it. Especially if you have many drives of different capacities.
If you were managing a farm of fileserve it would be different.
- as long as you have a couple of high cap volumes it is not anymore too much of a problem not to have everything in one volume

Also could you proritise some backup requirments amongst all those files.

If so, a solution could be :
1 - Take some of the less important and more "stable" stuff on the smallest drives and remove them (either you have a copy on DVD or a copy on a non-mirror remaining disk)
Those disks won't suffer and will be pretty enduring storage if they are properly stored away.
You could even mount them in 5 1/4 racks.
2 - This will give you some space to install new drives
3 - Get a big drive (or 2) and make a raid 0 out of it with you rolder drives. Either many RAID 1 with each of the small ones or make a Raid 0 of the later before (then raid 0+1).
Linux software raid is one quick and easy solution.
4 - Stuff all this in your home PC or something else to be sorted out and switch it on only when needing to access the files.

For about 100 USD you can get a 300Gb HD. Get 1 or 2.
This will give you a similar capacity of redundant storage. (multiply depending on what you can afford)
You'd be then left off with 600 or 300 Mb of storage that is unmirrored.
Put away of disks that you remove from the machine the stuff you have on CD/DVDs.
What is left won't probaly be that much.

NB : Install your OS on a small drive 20-40Gbcould be a good idea.

Posted: Sat Nov 11, 2006 8:09 pm
by geofelt
My 2 cents.
I am not a fan of RAID. Hard drives are very reliable these days. What RAID does is speed up the recovery process from a hardware failure. There are other types of failure that still need to be addressed. What would happen if you came home to a fire or other disaster that destroyed your PC and all of it's data. How about a Virus attack that erased your data?

First, decide what needs to be protected, and then get that backed up data to a safe offsite location. If you solve that issue, there likely will be no need for complex RAID schemes. Leave Raid to the 24/7 server guys.

Posted: Sun Nov 12, 2006 4:55 am
by kogi
Sell everything and start anew

1 x 500gb disk for windows + misc
4 x 500gb windows raid 5

Intel Allendale e6200
Gigabyte GA-965P-S3 (6 sata connectors)

Scythe Ninja

2gig of ram

Add a hefty seasonic PSU

large case like the stacker or the nine hundred

viola, all the points above

Posted: Sun Nov 12, 2006 6:13 am
by jjr
geofelt wrote: I am not a fan of RAID. Hard drives are very reliable these days. What RAID does is speed up the recovery process from a hardware failure.
Got a HD problem a couple of month ago with a brand new Samsung SpinPoint 250Go.
After 2 months of usage and probably a cumulated 30 hours running time it broke down. HD problem upon start-up : not recognised anymore.
Despite many hard tries never managed to have it even spin up !!
So much for the very reliable hard drives of today ...

Hadn't it be for the Raid 1 I would have probably lost everything (about 150 Go).

Finally changed it (was under warranty).
With a Raid 1, I could :
- transfer the files out,
- reinstall Linux (it was holding the sys partitions),
- rebuild the raid with the new drive,
- put all files back in,
and be up and running in a mere few hours.

Posted: Sun Nov 12, 2006 12:56 pm
by floffe
jjr: Raid 0 is striping == splitting the data over two or more HDDs (or partitions, if you use linux soft-raid). You're talking about raid 1, mirroring. If you lose one part of a raid 0 array pretty much all of the data is gone.

Posted: Sun Nov 12, 2006 8:47 pm
by jjr
floffe wrote:jjr: Raid 0 is striping == splitting the data over two or more HDDs (or partitions, if you use linux soft-raid). You're talking about raid 1, mirroring. If you lose one part of a raid 0 array pretty much all of the data is gone.
Sorry Raid 1 indeed : other message edited.

Save Yourself a few headaches

Posted: Sun Nov 12, 2006 11:22 pm
by waicool
Pick up a portable 500gb or 750 gb hard drive and start copying all of those crappy ide hd's onto it. Consolidate your files on one harddrive and then buy a backup for that. u r done. all portable good for any machine.

Posted: Sat Jan 20, 2007 8:45 am
by valnar
Sorry to dig up an old post, but my situation is the same. I've been following the Infrant ReadyNAS device for sometime because it's relatively cheap (compared to building a new system), RAID, low powered and has all the functions I need in a NAS box --- today. That's my only concern with buying a closed kit. I was ready to jump the gun on that until I read the newsbrief about Microsoft's new Home Server. Despite all the forthcoming complaints about giving more $$ to Microsoft, it looks like a damn good solution. This of course requires a PC, not something like the Infrant. 'Back to waiting.

I do have a couple old PIII machines at home. My only concern is the limited PCI bandwidth. If I buy a (e)SATA controller and a gigabit ethernet card, there goes all the PCI bandwidth. Since it won't be a local machine but only accessed over the network, it probably won't be a big deal if the speed of both dip a little during access. I just want to make sure saturating the PCI bandwidth doesn't cause reliability or corruption problems. I'm okay with speed suffering a little.

So now I'm back to square one. In order to make the PC upgradable in the future, I thought about buying an external SATA enclosure for 4-5 drives, but that also requires a PCI card to attach it to. That sounds the most "future proof" since I can put it on a better PC in the future with PCI-Express, built-in eSATA, etc. All my data is in an external box, not the PC itself. Then the only requirement is backup or redundancy, which either RAID (bad) or the new Microsoft Home Server (better) takes care of. I won't go into details here, but read up about the Windows Home Server and how it protects your data. It doesn't require RAID, and it may solve the problem of sticking a seperate USB hard drive on the PC for an offline backup.

From a home user perspective, I see these needs:

* Data has a backup so loss CANNOT occur.
* 24/7 operation is not required, or in some cases, not even wanted.
* relatively inexpensive
* non-proprietary. Data accessible in multiple ways or from multiple PC's if something dies.

Using RAID does not protect data from viruses, user stupidity or accidental erasure. As somebody mentioned above, all data is accessible at all times so there is no offline backup, unless you add that too. But then everything is redundantly redundant, with a price tag to match.

The Infrant NAS device is the same, minus the possible exception of viruses. A (X)RAID volume protects against hard drive loss, but a user can still erase a whole directory by accident, and voila, all your precious pictures are gone.

For a home user, RAID is not important. I agree. What is, is an offline or "impossible to get to" backup of your data that is somewhat hidden from the operating system, your own stupidity, or not attached to the PC at all, like a USB hard drive.

I'm still looking for the perfect solution, but it would probably be a PC with commodity hardware (standard motherboard, SATA drives, etc.) with a standard OS like Windows or Linux. If any part of it failed, the hard drives and data can be retrieved from another PC easily. The second requirement is a backup of some sorts that is not accessible by viruses, RAID failure, user activity or any other single component failure. Even better would be a time-delayed backup so that if you accidentally delete something, you can restore from a previous copy. RAID doesn't protect against that. Yes, a dead motherboard or bad power supply will prevent such a PC from booting, but at least the data is safe. I would assume most home users prefer that trade off. Businesses can afford both RAID and multiple backups. Most of us can't.

So what can we do? I dunno.... I'm back to waiting, but I'm looking forward to see what Microsoft is going to do.

Robert

Posted: Sat Jan 20, 2007 2:31 pm
by kogi
Freenas
Naslite
Openfiler

I haven't seen a NAS with built-in backup. Backup usually occurs on another box.

EDIT: It looks like Freenas and openfiler support data-sync and snapshots. I'll investigate further.

k

Posted: Sat Jan 20, 2007 2:32 pm
by kogi
meh doublepost

Posted: Mon Jan 29, 2007 3:35 am
by eztiger2
valnar wrote:Sorry to dig up an old post, but my situation is the same. I've been following the Infrant ReadyNAS device for sometime because it's relatively cheap (compared to building a new system), RAID, low powered and has all the functions I need in a NAS box --- today.
The Infrant NAS device is the same, minus the possible exception of viruses. A (X)RAID volume protects against hard drive loss, but a user can still erase a whole directory by accident, and voila, all your precious pictures are gone.
As an aside, and I'm sure you're aware, the ReadyNAS supports snapshots. It's not a great implementation (only hold one snapshot) but it would probably help you in your virus / accidental folder deletion problem provided you noticed in time.

We have ours set to snapshot once a day at midnight. Anyone deleting something by accident the following day can just grab it from the previous days snapshot.

Kev