I'm about to start a new Server build for home media storage. I've ordered an Areca 1280ML 24 Port Sata controller. If I will be running Ubuntu Server 10.04 LTS as my Server OS. Do I let the card configure RAID 6 or use Linux RAID?

I just google shopped that card and it cost about $1000 US.

There's no need to use software raid if you have that card. It will make all the RAID volumes you need and the overhead of running RAID will be on the card, not the server.

At boot the card handles the disk arrays, while the server just boots itself. You may see some speed improvement with the card too. You've bought a good piece of hardware, use it.

If you had a total of 24 bays available, Hardware Raid for all 24 and were starting fresh with a minimum of 12 x 2TB drives and assume using Raid 6. How many array's would you want to break 24 bay's? I had initially planned on 2 groups of 12, but that would be extremely expensive to grow all at once next time. Not to mention that I think 12 Disk in one Raid6 Array have performance issues or could be riskier.

In stead, I am open to 3 groups of 8, 4 groups of 6 or even 6 groups of 4, which ever makes the most sense from a usable and expandability perspective. I had $3K set aside for a new workstation. I will gladly use some of its budget for more disk if necessary to bump from 12 to 16 or whatever. I should also point out that going into the new Norco 24 bay chassis, I planned to reuse a MOBO with 4GB of RAM (mobo max) and a Core 2 Duo E4400 processor. Should I be looking to upgrade this MOBO to even a slightly higher end Desktop MOBO, Processor and 16of RAM? Would it make enough difference? Also, due to the fact that the NAS appliances I reference bellow boot from USB and can leverage SSD for additional caching performance. I have a 30GB SSD not doing anything. Should I use it for the OS install or mirror 2 smallish 7200RPM SATA drives. Is there any way Ubuntu could use this SSD to improve performance?

My worst case scenario budgeted usable data amount was based on 4 groups of 6. In the case of using 2TB drives, I think that is roughly 16 x 2TB(1.8 actual) or 28.8TB available? I understand there is a formula to follow for best performance and minimizing the impact of lost disk. I'm not describing it correctly, but what I mean is... I remember reading somewhere that when spanning(quite possibly the wrong term?) or sharing multiple Raid 6 array's, you don't always lose 2 drives for parity and the actual amount of usable disk space can grow. There was also a few good examples of what array's should look like. I wish I could find that article as I know I've massacred what it intended to say.

How many array's would you create to start? Will volume manager allow me to add new array's later to grow my usable disk space without having to move folders around? There has to be some way of growing without a complete rubuild otherwise I would have to assume software Raid wouldn't exist. Part of my thinking comes from unRAID, which I currently use. As long as the parity drive remains the largest in the array, you can add any number, size of disk and grow 1 disk at a time. FreeNAS and Nexanta also also allow growth through adding vdevs(I think). A bit more complicated than unRAID, but still doable, I believe? I have given up on all the other solution and want to make this a 100% Ubuntu Server solution.

My first stop was chapters for reading material as well any other book site for good reference material. I have googled, trolled IRC and posted in may forums and can not find any real instruction or advice for what I want to do. Most have simply stated the obvious, I (me) don't have enough knowledge to go the Ubuntu route. Regardless, I'm going to attempt it. Catastrophes ahead be damned! I have been warned! I've got Ubuntu Server running on a few DL380's in my basement rack and have rebuilt them as part of my own personal education. What I haven't been able to learn there, VirtualBOx on my laptop has been great. My prod box which really does nothing (DHCP, DNS, Internet Sharing, DDNS for my home LAN) is now GUI less, not even Webmin. The fact that I celebrate that statement should horrify a few of you!

Any guidance and feedback is appreciated. I know this is a long a post and somewhat abusing the intents of this forum. If I've pushed to far, tell me, laugh at, whatever. It's all good!

e

Firstly, anything simple that I could tell you about filesystems would be either deceptive or wrong (in some use cases).

However, from what you have said, the MoBo/memory combination sounds fine...but, you haven't really said what you want from it. you seem to be saying (reading between the lines, etc, etc) that you are building this for learning purposes and you don't have, eg, 100 users streaming video to and from the drive array. If this guess about what you are doing is wrong, please say what you are doing, as it will affect the answers.

In any case, you seem to be in a position in which you could build the system with the mobo/memory and look at how it performs; if the processor load or memory usage seem to be holding performance back, then decide to upgrade; you won't have lost anything over buying a new mobo now.

Could you use Ubuntu to make use of the SSD to improve performance? Well, a definitive yesno to that.

Firstly, from the size, I'd guess that its an older SSD, and, if so, the performance of that won't be as good as the performance of a device using the latest (sandforce?) controller.

Now, I have seen information from Sun on their ZFS filesystem and they show great progress in using a relatively small ssd to 'cache' (that's an oversimplification) a hard disk array and improve performance, and you could run ZFS on Ubuntu, via fuse. So, that's a yes, is it? Well, only in a very limited sense. A very no-like yes, if you like.

You need to look here and examine the numbers. ZFS really does take advantage of SSDs, but it starts off so much slower than Linux filesystems in most use cases that, while you can win with respect to ZFS-on-mechanical-hard-drives, more common Linux filesystems are so much faster that it isn't a win overall. And have a look here, as well.

Even that isn't the end of the story, though (not by a long way); in capability, ZFS is more like an ext fs + md + lvm, and, at the very least, there is a performance hit from lvm. So, at the very least, a comparison, ext + lvm versus ZFS would have been fairer, but raw the perf difference is large enough so that ext + lvm would still come out ahead (...but by how much? and would you be prapered to give up some perf for the manageability features of ZFS, which are superb (errr, imho)).

You could even consider Bcache, for example. But, when you've looked at the numbers, you probably won't.

Now BTRFS is a but immature, but is closer in capability to ZFS (and is closer in approach, too). you'll have seen than some of the benchmarks on BTRFS show it closer to ext4, but not really close enough; others are more even. In some use cases BTRFS is roughly as fast, in others it is really dramatic that BTRFS is slower. Which use case applies to you? Well, probably some combination of the two. And BTRFS has an SSD mode, so when it takes advantage of SSDs? Well, actually SSD mode, at this point, doesn't actually seem to speed BTRFS up, so that doesn't work, either, but maybe it will when BTRFS matures.

In any case, BTRFS is more 'manageable' than an ext fs on its own, so maybe you were happy to pay a performance price for that, particularly if in your use case CoW and zero cost snapshots (vaguely) make other parts of your workflow simpler.

There are some good articles on raid arrays (etc) on Wikipedia; that wasn't what you meant, was it?

Most people do software raid for either

• performance
• data security (some of which is illusory)

so, it would exist without a way of growing. Maybe you are thinking of MD?

If you really trolled, that would be unsurprising; did you mean 'lurked'?


If you take this statement into account when planning a backup policy, then fair enough!

That's not nothing...for a small number of users, probably doesn't overwork the box concerned, though, if that's a problem.

I don't see this as abusing the intent, and, as you say, it's all good. The one comment that I would make is:

What you have done there is buy a decent controller card; most of the stupid threads that there are on raid are from people who have determined to use the 'free' raid that comes with some MoBos or the cheap add-on raid cards. This (rather than using the MD driver, with such hardware) is a risky move, from the point-of-view of data security. Combine this with drives not intended for RAID usage (there are rumours that WD deliberately ham-string some of their good-value-by-the-GB drives, because they want to push people to their enterprise-spec drive, but I think it may be accidental, rather than deliberate, and this may be something of a conspiracy theory) and intermittent power, and you have almost certainly have created a problem.

Oh, and if you go for 4k sector drives, be sure that you know how you will deal with the problem.

Salasi,

Thank you for your comprehensive reply. When I initially started working on this "project" 6 months ago, I was going to use the biggest tower pc I could and run windows home server on it. At the time, that appeared to be the easiest, cleanest option. The only concern I had was that Vail will be released in early 2011 with not direct migration path.

Since my initial investigation began, I have learned many things. Most importantly, there is a huge world out there beyond Microsoft AND I love it. Although I have not been able to get programs like my Home Design software suite running cleanly on Wine or satisfactorily in a VM. I have been able to remove MS significantly from my personal life. I am in no way bashing them, just stating I've tried to find a different way of doing things. Also as a point of clarity... I have been "lurking" (not trolling) in the IRC channels and posting questions and engaging in conversations with good and bad results. I'm in Atlantic Canada and have managed to make a few friends around the world. I can't even make new friends down the road from our house! I naively believed IRC ceased to exist when I left my parents house 18 years ago. I thought it and "Gopher" both disappeared around the same time.

Further definition around my intentions. This is not meant to be a learning project or continuing education as I may have implied. Like many others, I have growing media collection as well as data I've accumulated over the years that I want to have accessible. I'll touch on backing this up in another paragraph. I am currently using a 2 year old HP Pavillion Tower + PCIe Port Multiplier + 8 bay Sans-Digital External Enclosure and unRAID. This configuration allows me 3 x 2TB drives in the Tower, 8 in the external Enclosure and I boot unRAID from a 4GB USB Key. This solution has worked well and I paid $150USD for the 22 HD solution. I do find it slow for uploading etc... Not slow from a play-back perspective.

I have a 42U Rack in my basement with all networking, computing and UPS contained within in. My current NAS sits on the floor of this rack. I want something I can mount and get higher than 2" off the floor. I have a Norco RPC-4224 boxed up and waiting for me to figure out next steps. My drives are all inexpensive Samsung 2TB drives. I have a total of 12 right now. 5 of which are still unopened. Whether I made the right choice or not, I also ordered a new MOBO, Processor and 8GB of RAM. If need be, I can bump up to 16GB down the road. Additionally, as happens with NewEgg from time to time, they split up my order and the Areca card had not shipped when I checked yesterday. Upon further reading, I found that the Areca 1680ix was faster and more current. Rightly or unnecessarily, I upgraded to this card.

About my requirements... This is a home solution. I do however expect that like any other SOHO solution, my data will be safe. A lot of my ideas came from this post. We have dedicated HTPC's in our Theatre Room, Family Room and Master Bedroom. Both our kids of Win7 laptops, My wife has a Win7 laptop, I have a work laptop, personal laptop and a new tower for ripping my Blu-Ray collection. I also have an iPad I'm hoping to find a streaming solution for when I'm feeling lazy. It's the Blu-Ray collection that will drive my growth. I am expecting an average of 25GB per, for a full ISO backup. Over time, and using Avatar as an example that number will shift to 50GB per movie. I have about 450 blu-rays and 250 DVD's I need to get through now.

Now that you've got more history than you require. Let me try and clarify my intentions. Upon further reading on my Areca controller. I find that it has the ability on its own to grow my array disk by disk. Whether I go this route or simply pool a few Raid 6 arrays together using LVM, I'm not sure yet. I do intend to use Ubuntu Server as many have pointed out $1,300 for the Areca card is wasted for simply growing my SATA drive count with FreeNAS or Nexanta. As much as I find ZFS and soon BTRFS intriguing, I don't see why I need it - yet. From a performance perspective, I have not been able to max out unRAID even with 3 simultaneous files playing. It could be that my HTPC and Storage are all on a GB switch? For the OS, I will re-use 2 320GB, 7200RPM SATA drives I'm not using. They are much larger than I need, but free.

The last piece I need to add is backup. I'm really struggling with this. Initially, I wanted to back up all my data. Especially based on the ripping time I'm looking at. I can't afford that however. Like some others have done, I may need to simply try and prioritize my data based on its importance to me. I was originally going to sell my HP multimedia box and Sans-Digital Enclosure. Now, I'm thinking I may keep'm both for to use for backup. Not sure what OS to use? Heck, I even contemplated hanging the enclosure off my new Norco server MOBO. However the two problems I see is that the enclosure would still end up on the floor and that my backup ends up situated physically within the same structure. I don't think either of those could be considered advisable. Still, backup is an area I still need to come up with a viable solution for. I wish cloud solutions were more affordable. I have a 70Mbps fibre connection coming into the house. Surely my ISP could handle incremental backups once my initial data is taken care of...

I'm hoping this adds some clarity.

e