What NAS HDD speed...

[Fayd]

I'm sure too low is the answer. Really, anything not in the hundreds is too low, and even then, he would only just be reaching a significant sample size, except for drives that have a high failure rate (Segate 7200.12 now, IBM 75GXP of ages past). I've had three Samsungs fail, one in a RAID, one still within the RMA period. They must be terrible drives.

I have had really bad luck with newer (.11 and .12) Seagates, by way of infant mortality, and a nice firmware bug, so am avoiding them (by Newegg reviews, I'm far from alone). Otherwise, it's hard to tell, until enough people have enough experience, at which point the drive series will have been replaced. IMO, the best policy is a backup drive, of a different drive series than the drive it is copying from.

If using the NAS as a file server, with no meaningful amount of random R/W, faster drives could help, but there isn't a good way to know how much, unless someone benchmarks the difference. Given the RAID 5 performance in that link, it looks like the NAS is able to do a good job of prefetching and buffering, for sequential transfers. In that case, there would probably not be a difference. But, when you start dealing with smaller files (with RAID 5, even single-digit MB might be small enough), the difference might start becoming apparent.

IMO, if you can afford a $700+ NAS, you might as well spend a little extra and be sure you won't have drives that slow you down. That said, if you want to try slower-spinning drives, Samsung's seem to perform the best in random loads, of those slower drives.

i've had 3/7 samsung drives i've had fail within the first 2 years of ownership.

i don't think they're terrible drives, i think i just have bad luck. i've since switched to hitachi drives.

 

[imagoon]

In terms of speed (perfect controller):
RAID0 - 200 MB/Sec
RAID5 - 150 MB/sec
RAID10 - 100MB/sec
RAID6 - 100MB/sec

Can you explain how you arrived at these numbers? It certainly doesn't match my experience. That or I am not understanding what you you are saying.

Assuming 100% random read with each disk delivering 50MB/sec while 100% random those numbers would be theoretical max read rates but once you start adding writes and things like stripe sizes and overhead it changes drastically. RAID 5/6 are 'read intensive' during the write pass plus the XOR time and as such tend to be quite 'slow' as the write rate increases.

 

[Zebo]

Damn. I should be buying lotto tickets then. I have had this issue more than once. Typically with larger capacity disks. Old 15k 146GB drives chug for 10 years, 600GB 15k SAS disks are accumulating errors.

It is statistics at work. As the disk sizes grow (and sector count increases), the odds of a read error increases. When a disk is failed out and offline, all it takes is one bad sector during the rebuild to bomb an array.

True bit rot etc, that's why a lot of people are moving to Suns ZFS. But it's still really rare for two at once.

After doing a bit of research on my own it seems BYO is the way to fly. (Thanks for the links/stats guys)Not only do you get superior hardware used in higher end prosumer NAS's like DS1511+ for 1/3rd the price you get to use the file system you want and a superior file system built from ground up with error control in mind. I think I'm going to sell DS411j and move on keeping disks and adding 4 more with ZFS. Disk speed seems irrelevant as GB pipe is smaller than 5400 RPM s can deliver. That is the weakest link and wireless even more which most clients use.

 

[imagoon]

True bit rot etc, that's why a lot of people are moving to Suns ZFS. But it's still really rare for two at once.

After doing a bit of research on my own it seems BYO is the way to fly. (Thanks for the links/stats guys)Not only do you get superior hardware used in higher end prosumer NAS's like DS1511+ for 1/3rd the price you get to use the file system you want and a superior file system built from ground up with error control in mind. I think I'm going to sell DS411j and move on keeping disks and adding 4 more with ZFS. Disk speed seems irrelevant as GB pipe is smaller than 5400 RPM s can deliver. That is the weakest link and wireless even more which most clients use.

Actually so far my group has been a bit 'fearful' of ZFS at the moment because Oracle took over Sun. Granted I should mention at this point I have not touched / do not use consumer NAS's... Everything I have is Enterprise / Datacenter. A 5400 RPM drive would not be able to saturate the links we have going to our SAN gear.

Data throughput doesn't mean much anyway. It is all about IOPs which the 5400 drive would fail hard on.

 

[LANMAN]

For the cost difference (which isn't much) go with the 7200 drives.
I don't use 5400's for anything anymore; they're just too slow.
As for raid failure, I agree about the lotto ticket -- I've seen RAID5 arrays fail, but out of the hundreds I've managed, only 2 were unrecoverable. (Mostly because the end user who was in charge of the managing the array, didn't do proper maintenance)
My motto -- the network is only as good as the administrator behind it.

--LANMAN