[Discuss] NAS: lots of bays vs. lots of boxes

[Rich Braun]

As drive capacities increased, transfer speed also did. (You have updated your motherboards to 6G SATA, right?) I posted here not too long ago that I'd suffered a triple-disk failure, which forced me to buy into the current crop of magnetic media rather than wait & see what the SSD market looks like in 2017.

I maintain a pair of servers, containing (more or less) copies of the same ~10TB of data. Seagate lost my trust once & for all so what I bought was a half-dozen more 4TB and 6TB HGST units so I could discard the last of my Seagates.

I use software RAID5 just as I described in a lengthy writeup here about, oh, 10 years ago. There are now 4 drives in one server and 5 in the other. To prevent RAID sync from taking too long on any given operation, I divide up each drive into 1TB partitions and combine 4 or 5 of those into a volume group. I have 5 such volume groups now. 

Rebuilding a volume group takes a number of hours: overnight, vs the old days of a couple hours to sync say 100GB. No big deal. Should I go for RAID6? No, I run two mirrored servers and several backups by 3 different methods.

I bought the first of these two servers as hardware RAID1 with a pair of 30GB drives in 1999. Sixteen years later, my volume-management practices haven't changed (at least, since ditching hardware RAID) and don't take appreciably longer as storage requirements marched upward. The main thing I've added recently is LUKS encryption; the main thing I've discarded is GlusterFS.

Why so much data? Music, a couple thousand DVDs, a quickly-expanding library of family photos and videos, a lab for virtualization / containerization. Linux has been my main hobby and work pursuit for 22.5 years.

What's going to upend these NAS RAID technologies in the future are two things: cheaper SSD (or the end of HGST's high-quality magnetic drives), and cheaper cloud solutions/network capacity. The latter might still be 15 years out.

-rich