New Network Attached Storage suggestions
#46
darkscout Wrote:ZFS FTW.

Starting to look for an alternative since OS is dead.
I get this sentiment a lot from people, but if you're using OpenSolaris as just a file server basically, I don't see what's wrong with staying behind the curve for a while until a viable OS shows up (ala Illumos) or it's cheap enough to migrate all your data to a new setup. After all, when billion-dollar companies can stick with old Linux / Unix systems like SCO Unix and mainframes (despite no support!) there's probably a bit of room to stick with OSes that are supposedly dead in the water like OpenSolaris.

I wouldn't have any hope of FreeBSD effectively "catching up" to the OpenSolaris ZFS implementation. A great deal of the ZFS architecture is built upon Solaris constructs that presents challenges to porting to alternative kernels ugly. Look at all the memory adjustments you need to get ZFS working well on FreeBSD, for example.

I'm going to stick with an old, moldy OpenSolaris system for my 12TB disk setup until:
1. btrfs is mature and trustworthy enough to hold all my data
2. Illumos or another OpenSolaris successor can safely run ZFS v17+
3. Drives get large and cheap enough for me to just run a single disk in RAID1
Reply
#47
Yeah I had a zfs array myself with 8 drives in raidz2 the only reason I don't now is because I decided I wanted to set the server up as a second desktop running server 2008r2 hyperv. Which was fine for accessing the raid but drives would randomly drop out and eventually the array was screwed. The problem I had with zfs is the annoyances of having to setup acl's. Although I see there use at home it's not needed. And when compared to Linux md it's not as safe in terms of repairing problems or recovering and there is occluded the problem of being limited to opensolaris or zfs fuse.

Now that opensolaris is dead and my array was dead I decided to go with ubuntu raid5 on 3x2tb disks with easy online expansion. Would of went with centos but for the extra stuff like sabnzbd ubuntu is a better choice. And eventually I will move to btrfs unless oracle screws that over to.

And as for running a fileserver with disks that can spin down. I use it during the day for things and download at night. So I went for the a cheap energy efficient and setup with wd green drives. This IMO is a far better solution then running multiple computers for different services. And I think it's possible to actually spin down the array although I'm not positive.

Please excuse my typing as it was all done on the iPhone. :-)

I should also add I suggest staying away from operating systems that are built on dead projects. Without the backing of sun or oracle I can't see solaris being useful I'm a server situation unless you actually pay for it. It's a shame really. Opensolaris was great with suns support.
Reply
#48
The reason I'm looking for a replacement is that I want to upgrade to Xen4. I'm also looking at OracleVM as an alternative. But given that they own Solaris now, I have no clue why OVM is built on a fork of RedHat (Which can't read ZFS in the context of the Dom0).

I looked up md, and it doesn't seem to have near the repair or checking capabilities of zpool. It's also 'limited' to FreeBSD, Nexenta, and (hopefully soon Debian GNU/kFreeBSD).

And OpenSolaris may be dead, but Solaris sure as heck isn't. Oracle just changed the way that they're going to roll out changes/source code. You're more then welcome to download solaris for free.
http://www.oracle.com/technetwork/server...index.html

You'll get no support and the GUI looks like something that stepped out of the 90s, but under the hood it's all new, including the latest ZFS.

-
Newegg has a sale through 10/2 for a: ARK 4U-500-CA Black 4U Rackmount Case 3 External 5.25" Drive Bays w/free HighPoint Rocket 620 PCI-Express 2.0 x1 Low Profile Ready SATA 6.0Gb/s Controller Card.

Toss in a 5 in 3 for the 5.25 bays and you could have yourself a nice 11 drive case for relatively cheap.
Reply
#49
joel_ezekiel Wrote:And as for running a fileserver with disks that can spin down. I use it during the day for things and download at night. So I went for the a cheap energy efficient and setup with wd green drives. This IMO is a far better solution then running multiple computers for different services. And I think it's possible to actually spin down the array although I'm not positive.

In a 3-disk RAID5 solution, the disks don't spin down unless there is no usage. This is because the data is striped across two disks with parity on the third. Not only does this mean that you are using more energy (less of a concern for me), it means that the life of the hard drive will be shorter.

RAID5 is an enterprise solution because it is good for cases where cost and length of service are far less important than 24/7 availability. Traditionally, you setup RAID5 after filling a server with disks and then make the space available for applications. Very little change happens until you replace the storage server.

In the scenario I described earlier you would achieve better results even on a 3 disk array. Writing to disk only involves 2 disks (one content drive and the parity disk). If you let it manage the distribution of content across the drives, each drive may effectively have half the usage it would have in RAID5. In arrays with more disks, the advantage is even greater. This means you can actually get more life out of several small disks than you can fewer large disks.

It's all about meeting your goals. For me, I would rather have a compute server (AMD based, about $350-$400) and a storage server (about $300 without disks) that can each be tuned to be the best at their jobs. The storage server takes almost no power because it uses a low power chip and spins down drives aggressively. More importantly (to me), it uses very inexpensive and extremely stable components because the most money will be spent on the disks. Nothing on the storage server is compute-intensive. The compute server is stripped down for speed with faster processor and memory, a small fast hard drive and a gigabit ethernet connection to tons of storage. I'd rather spend $700 dollars on two machines spec'd well for their tasks than $1000-$1500 on a machine that performed almost as well at both.
Reply
#50
darkscout Wrote:The reason I'm looking for a replacement is that I want to upgrade to Xen4. I'm also looking at OracleVM as an alternative. But given that they own Solaris now, I have no clue why OVM is built on a fork of RedHat (Which can't read ZFS in the context of the Dom0).

I looked up md, and it doesn't seem to have near the repair or checking capabilities of zpool. It's also 'limited' to FreeBSD, Nexenta, and (hopefully soon Debian GNU/kFreeBSD).

And OpenSolaris may be dead, but Solaris sure as heck isn't. Oracle just changed the way that they're going to roll out changes/source code. You're more then welcome to download solaris for free.
http://www.oracle.com/technetwork/server...index.html

You'll get no support and the GUI looks like something that stepped out of the 90s, but under the hood it's all new, including the latest ZFS.

-
Newegg has a sale through 10/2 for a: ARK 4U-500-CA Black 4U Rackmount Case 3 External 5.25" Drive Bays w/free HighPoint Rocket 620 PCI-Express 2.0 x1 Low Profile Ready SATA 6.0Gb/s Controller Card.

Toss in a 5 in 3 for the 5.25 bays and you could have yourself a nice 11 drive case for relatively cheap.


You might hate unRAID, but you will find some useful information in their recommended build information here. It outlines cases, motherboards and SATA controllers with expected prices and performance. For instance, I can get a Cooler Master 590 case with 9 external 5.25 bays, which can be turned into a total of 15 hot swap drive bays for tons of future expansion.
Reply
#51
lifespan of the disks i dont consider a factor, although a agree with mostly what everyone has said since my last post. I do agree that unraid is good for power consumption and the ability to read from only one disk and spin down the rest, this is why i tried it out myself. But the limitations it brings and the use of reiserFS made me unhappy, speed again not that important but nice to be able to saturate my gigabit when i use it as storage for everything. As for solaris, solaris 10 has been free for long time and the next version of solaris (OpenSolaris) which is no longer open. I believe is a trial now and as far as i can see no more opensource updates for zfs.

And when i said linux md is better then zfs for repairing, yes zfs is better and safe to use but if something goes horribly wrong its far easier to recover data from linux md array imo.

Anyway i do certainly like the idea of spinning down disks that you do not use with unraid i just really wish there was a more open solution so you can setup a server properly using it. But with green drives, and i plan to build up to 2x6drive arrays and replace the smaller array with whatever is the best value at the time that way atleast limiting the amount of drives im using and i can spindown the unused array. Who knows maybe there will be a better solution in the future for doing everything i want with the one OS on the one machine Big Grin
Reply
#52
joel_ezekiel Wrote:lifespan of the disks i dont consider a factor, although a agree with mostly what everyone has said since my last post. I do agree that unraid is good for power consumption and the ability to read from only one disk and spin down the rest, this is why i tried it out myself. But the limitations it brings and the use of reiserFS made me unhappy, speed again not that important but nice to be able to saturate my gigabit when i use it as storage for everything. As for solaris, solaris 10 has been free for long time and the next version of solaris (OpenSolaris) which is no longer open. I believe is a trial now and as far as i can see no more opensource updates for zfs.

And when i said linux md is better then zfs for repairing, yes zfs is better and safe to use but if something goes horribly wrong its far easier to recover data from linux md array imo.

Anyway i do certainly like the idea of spinning down disks that you do not use with unraid i just really wish there was a more open solution so you can setup a server properly using it. But with green drives, and i plan to build up to 2x6drive arrays and replace the smaller array with whatever is the best value at the time that way atleast limiting the amount of drives im using and i can spindown the unused array. Who knows maybe there will be a better solution in the future for doing everything i want with the one OS on the one machine Big Grin


In your situation (wanting one server to rule them all), I can agree that unRAID is definitely not the answer. I would argue that Ubuntu or RedHat would be a much better solution long-term, though. OpenSolaris' biggest fault is that Solaris was always an "against-the-grain" distribution. For a general purpose server, I tend to want something with nearly universal availability of packages. OpenSolaris easily falls into the "not quite open enough" category that unRAID occupies. While I am comfortable with a niche product being so tightly bound to one vendor, I would never consider a general purpose server with the same issue.

But it makes for an interesting discussion, nonetheless.
Reply
#53
That's the thing that disappoints me with opensolaris. I believed it was better then Linux. The filesystem far superior atleast until we get btrfs and with most of the development done by sun with the backing of sun it had real potential. It was already possible to get mostly anything running on it without great difficulty. But thanks oracle for ruining that for us.

Without corporate backing any further solaris based distros will be just another distro like the other 500 Linux ones already there however have no developer support from other projects. :-(
Reply
#54
Define "corporate backing". Nexenta is a real company. Illumos is going to pick up with OS left off. Oracle is still backing Solaris, they're cutting OpenSolaris.

ZFS is licensed under the CDDL. Oracle didn't say they would stop releasing Sun updates. They're changing the model in which they do it.

Instead of developers having access to the 'latest' code like they did, Oracle is going to wait until after the corresponding version of Solaris is released.

And who knows what will happen. Oracle is a major source of funds for brtfs... they now also own ZFS. They could somehow just merge the projects and release it as GPL.

And I've never had any problem getting packages for OS. It's not a standard desktop distro, so of course you can't get EVERYTHING under the sun, but I'm running SABnzbd on it
Reply
#55
joel_ezekiel Wrote:Well I hope this helps people debating if they want to use unraid or not. Basicly yes it's acceptable if you want slow speed cheap nas. But no if you want well performing server that can do many things.

You make it sound like Unraid is not up to the task.

Unraid works PERFECTLY as a media server. Not once has its "slow speed" caused a playback stutter or a dropped connection. Heck, the Samba in the Unraid distro is so tweaked it BLOWS AWAY the default Samba in Ubuntu for performance.

The only reasons to not use Unraid are:

1. You want your server to do more than Unraid.

2. You want to simulate a business network environment in your home (I couldn't guess why, home business maybe?)


But please don't act like Unraid is unacceptable for a media server. In fact, its probably the only RAID out there DESIGNED for that task. The advantages it offers over Linux RAID or ZFS (such as spinning down disks, easy to grow arrays, the ability to mix and match drives) are basically the exact advantages that consumers like those on the forum could want....

Reply
#56
poofyhairguy Wrote:. . .In fact, its probably the only RAID out there DESIGNED for that task.

Just a small flaw there. It isn't a RAID implementation. RAID means something very specific. If it were a RAID implementation, it would not be called unRAID (or they'd just be really dumb to call it that).

In the interest of being fair, unRAID would fail miserably at tasks for which RAID (with hardware controllers and a large memory cache) was particularly well equipped:
  • Constant small file access
  • High Simultaneous User Count (dozens to hundreds)
  • Applications where number of writes approaches or exceeds number of reads (i.e., logging, online backup for large networks)

You just have to remember that these are vastly different use cases.
Reply
#57
GJones Wrote:Keep the other computing tasks on a separate server unless it is very media-intensive. It will be less expensive and more stable that way in the long run.

I 100% agree with this. Sickbeard, Sabnzbd, and coachpotato are all works in progress and they need constant updating to keep up with the features until they hit some sort of stable level (the day when they are all one program).

If you just have one server, then you are constantly fiddling with it which leads to downtime. This is just three programs, you throw in a few more for the server (for me its also Air Video, MythTV, Orb) and suddenly you are faced with a situation where it requires almost constant fiddling to keep features from breaking because you have obsolete versions of the software.

So basically in that scenario you are constantly fiddling with your server that is supposed to be the stable backbone of your XBMC setup. No thank you, power and parts aren't that expensive, so personally let my servers only be servers.

For all the other stuff (Air Video, MythTV, Orb, Sickbeard, Sabnzbd, and coachpotato) I have turned my bedroom HTPC into a quad core monster that play XBMC while it does all these other tasks. You would be amazed how quiet you can get a quad core to be if you buy the right heat sink and silent 120mm fans. In this solution the box I am constantly fiddling with (and therefore breaking) is my least used HTPC, and since I would have a bedroom HTPC no matter what this doesn't add a box, it just reassigns the tasks...

Reply
#58
GJones Wrote:Just a small flaw there. It isn't a RAID implementation. RAID means something very specific. If it were a RAID implementation, it would not be called unRAID (or they'd just be really dumb to call it that).

In the interest of being fair, unRAID would fail miserably at tasks for which RAID (with hardware controllers and a large memory cache) was particularly well equipped:
  • Constant small file access
  • High Simultaneous User Count (dozens to hundreds)
  • Applications where number of writes approaches or exceeds number of reads (i.e., logging, online backup for large networks)

You just have to remember that these are vastly different use cases.

Agreed 100%. I consider Unraid to basically be RAID 4.5, which is why I called it that. It uses RAID 4's single parity disk, but the data is not stripped like RAID 4.

I agree to its limitations above, with the caveat that NONE of those things matter for a media server....

Reply
#59
poofyhairguy Wrote:I agree to its limitations above, with the caveat that NONE of those things matter for a media server....

True. I just do not want someone less experienced with these issues to walk away thinking that it was a generalized solution to replace RAID.
Reply
#60
GJones Wrote:True. I just do not want someone less experienced with these issues to walk away thinking that it was a generalized solution to replace RAID.

Good call.

Reply

Logout Mark Read Team Forum Stats Members Help
New Network Attached Storage suggestions0