Bay pride: WD slips out FOUR-DRIVE network FILE STASH WD has fixed the limited space and lack of RAID on its 1-bay My Cloud box by introducing a 4-bay one, the EX4. The first My Cloud box, providing data access to all your home/small biz network, was powered by a dual-core ARMv7 processor running Debian Linux. The single-core, 2GHz Marvell 88F6282 processor inside this new 4-bay …
For a home user who uses this for backups/storage/video then the downside of RAID is the rebuild time when disk failure occurs.
Sadly the now defunct Windows Home Server 2008 system had this just about right for a home user - every file (bar backups of client machines) was duplicated onto two physical hard disks. You get a disk failure and there is no downtime - plug a new disk in and the OS would sort it out in the background.
RAID is no doubt technically better but the old WHS 2008 was much better from a user perspective - even if the system disk died you could read the data from individual disks for recovery just by connecting them to another machine.
MS did some things right - then binned them :P
You get a disk failure on a hot-swap-capable RAID box and you do exactly the same. The operating system rebuilds the array in the background while you continue to use the degraded array.
Don't know which of the various home array boxes are this capable but you can build your own which is, based on a PC, Linux software RAID, hot-swap disk cage(s) and a hot-swap-capable SATA controller.
Yes but try streaming/backing up to/from it while it is rebuilding the RAID array. Painful is an understatement for streaming and backups won't happen until it rebuilds. WHS 2008 (even on an Atom) copes with that as its a lot less cpu/disk intensive to copy files in the background to a seperate physical disk than it is to rebuild an array.
What's the rebuild time going to be on a RAID5 2TB disk array on this? I'd guess maybe a day or so? Thats been my experience on similar stuff.
RAID isn't always the right solution.
On Linux you can throttle the rebuild megabytes/second rate, if you are willing to trade a slower rebuild for a greater responsiveness during the rebuild. It also prioritizes file I/O over rebuild I/O. If your workload is still seriously impacted, it suggests to me that the hardware is only marginally adequate even under ideal (non-rebuild) conditions.
Rebuild time of a RAID array with no overlying filesystem activity is simply the time to read N disks and write one. If the controller is capable of supporting all N disks streaming in parallel, for a typical 150 Mbyte/s 2Tb drive that's about 13000 seconds or ~4 hours flat-out, 8-12 hours if throttled to allow overlying filesystem activity. A whole day would suggest a bottleneck in the controller or the CPU. Soft RAID on an Atom CPU? What's the maximum rate it can calculate XORs? OTOH a decent Core-2 quad can outperform 4 SATA disk drives without breaking sweat (and still has three cores left for other work).
The number of I/O operations needed to copy a whole filesytem as files (indeed, even to read a whole filesystem as files) exceeds the number of operations needed to rebuild the underlying RAID block device (apart from the case where the filesystem is mostly empty, or partly empty with the rest made out of a small number of very large files). Also rebuilding does sequential block reading, with only one-cylinder seeks. Accessing a filesystem will tend to do random seeks in larger numbers. So there's no fundamental advantage in doing mirroring at the file level. (I'm not saying that there can't be an advantage when particular implementations are compared).
My sixth sense tells me there are very few posters here who actually understand what RAID is, nor how the different types of RAID work or that there ARE different types of RAID. More so, that there are different implementations of the same RAID level.
Firstly, drive mirroring is actually RAID 1. Saying RAID is bad and mirroring is good, only shows your ignorance.
RAID 5 is not "RAID". Look it up, RAID covers a fair range of options. The important ones notable on domestic NAS arrays being 1, 5 and 6. Might be worth looking up 0 as well, even though it doesn't form any redundancy, it can improve performance.
Just because your shitbox PC can't access files during a rebuild, doesn't mean ALL implementations are like this. My Synology box (and my QNAP for that matter) are fully functional for file availablity, and built-in software availability during rebuilds. In fact, if you weren't actually looking at it (or listening to the alarms going off when one of your drives failed), you wouldn't even notice aside from the slower speed response.
I've been goofing off with RAID for two decades, have NEVER come across availability issues during failures or rebuilds (again, aside from speed issues).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
