>>46841984
>Nay.

Build your own. I'm doing that myself. But if you really want an ultimately cheaper all-in-one package, you can go with that. It's cheapest if you already have a computer laying around that you don't use and can install/configure FreeNAS with.

Honestly NASes are shit for storage speeds because they'll get bottlenecked by the network. Until they have 10GbE (or someone comes up with ethernet bonding that doesn't suck) they'll be slow as shit (especially a RAID NAS).

>>46842030
HDD's can only write at around 150mb/s, a "standard" gigabit network has a bandwidth of 125mb/s (1 gigabit). with a decent switch you're right.

>>46842073
It's a little lower when you consider the overhead from layers 1 to 7. You'll probably max out at about 115MB/s.
But if the NAS/SAN is doing RAID, then the theoretical max is 150MB/s per drive. The real-world throughput is a lot lower, but still going to get significantly bottlenecked by the network.

>>46842112
Link aggregation may be an option, really I'm looking for around 100mb/s sustained, which should be doable.

>>46842134
The problem with link aggregation is that the only implementation of it that can actually use multiple physical links to improve the bandwidth of a single TCP/IP (or UDP for that matter) connection is Linux's balance-rr style (I'd guess BSD has an equivalent). Most forms of link aggregation refuse to split up packets that appear to be from the same connection (sometimes even from or to the same host) because they don't want to fuck up the order of the packets.
As a result it's pretty much useless except when you're going to have many concurrent connections.

>>46842158
ah, I was unaware of that. even without it though the network should be able to push 100mb/s though

>>46842012
freenas requires ECC memory, which means that you have to go with a server or workstation grade motherboard which increase the cost significantly, hence the reason I'm considering the nas unit.

>>46842199
If the ECC requirement isn't something that can be bypassed, just install whatever Linux distribution and install samba/targetcli/whatever other NAS/SAN software.

>>46842208
most, if not all use memory as a disk buffer, so couldn't ram errors still affect data integrity? wouldn't you still want ECC ram regardless if you were using ZFS with freenas or samba/afp with an ext4 file system on another linux distro?

>>46841984
>Netgear ReadyNAS 104 4 Bay NAS
Does these stuff have ECC ram?
>512 MB ram
Is this even enough for ZFS?

>>46842249
>RAM errors
Those basically never happen to the point where you don't need to worry at all
If it did happen, then it would be no worse than it happening on your desktop. Does your desktop have ECC ram?

>>46842255
I'd assume its not using a ZFS filesystem. and 512mb would not allow for memory to be used as a buffer, hence the lower performance.

>>46842266
I know they're rare, but they can happen. and I'd like to ensure the reliability of the server and file system.

>Does your desktop have ECC ram?
No, it doesn't require it.

>>46842249
ZFS is a special case where 1 error can corrupt everything on your disk and still keep going.
And because it keeps going your backups are going to the shitter as well. The process can take months.

>>46842288
so if I wasn't using ZFS (say, ext4) and there was a RAM error with the buffer, what would happen? would the error be detected?

>>46842288
Isn't this a bug of silent error correction of the drives by ZFS? I would think that btrfs is vulnerable to this bug too.

>>46842310
Nope. It just wouldn't spread the error in your FS and it would remain local.

>>46842325
is there a way to check that a ram error hasn't occurred when copying files to the server? (if it was ext4 or any other type of partition)

>>46842310
It would be just that 1 error that went bad. if it happens to be your filesystem you'll lose it, but you can still recover all data from it. Other then that I don't you'd mind a different pixel somewhere.
data on conventional filesystems don't make hashes and don't depend on each other.
ZFS is more of a expensive work solution. And it's good at that. Just don't use it if you have no idea what you're doing. People would like you to believe it's fine to run without ECC and without physical access to drives(Virtual machine etc)

>>46842343
I don't think so. Also RAM errors could happen everywhere, not just in your FS cache. It could happen in your image editors cache where you save an image, there is no way that you could detect that. It could happen in the running code of the FS, it could go berserk and destroy your FS (most likely it would cause a kernel panic tho).

>>46842390
could you verify it by its hash? like md5 or something?

>>46842411
that's the solution and problem with ZFS
it hashes hashes upon hashes etc
1 hash wrong and everything falls apart
use ECC and it's the best thing ever
get it?