[gentoo-amd64] Re: New install, /var on tmpfs

[Duncan]

Volker Armin Hemmann <volker.armin.hemmann@××××××××××××.de> posted
200802061341.37169.volker.armin.hemmann@××××××××××××.de, excerpted below,
on  Wed, 06 Feb 2008 13:41:37 +0100:

> and the more spindles, the higher the chance for a fatal disk crash.

Well yes, but that's sort of the point of using RAID-6 instead of RAID-5, 
and once one gets over 8-10 disks, RAID-10 or RAID-61 (? or would it be 
called 16, duplicating each disk, not the entire RAID-6 ?) becomes 
practical.

Actually, today I'd probably implement it using a couple of those 
external 5-way port-multipliered SATA boxes, connecting each to the 
system via single SATA-300 cable, which should handle 5 disk's disk-rate 
I/O reasonably well without bottlenecking.  Connect each cable to a 
separate SATA-300 card on separate PCI-E or PCI-X channels, avoiding 
bottlenecking there, and the bandwidth should be reasonable.  Because the 
external disk arrays have their own power supplies, the one in the 
computer itself wouldn't have to be such a honker, I could run low noise 
supplies in all of them, and the computer case could be smaller too, as 
the only drive in it would be a DVD burner.  It'd be three smaller, 
quieter, more portable and easier to work with boxes, in place of the 
single full tower case, quad internal hard drives plus DVD burner, I'm 
using today.  The dual 5-way drive cases would give me space for ten 
drives, of which I'd probably make two hot-spares, leaving an 8-way 
RAID-6, two parity, six data stripes.  Large write performance, the real 
killer now, would be much better, and because I'd have two hot-spares 
ready for the system to bring online if necessary, total reliability 
should remain comparable to now, 4-way RAID-6 but without the hot-spares.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-amd64@l.g.o mailing list