You are not logged in.
So, I've got a set of 6-750GB SATA drives (Seagate, if anyone cares) that will all be connected to the onboard SATA controller of my Intel Seaburg chipset motherboard. My goal for the system is to provide a stable workstation for doing GIS work and coding in Python\C and perhaps also photo cataloging and storage. The workstation will be my primary system, as I don't have two desktops. I've been working on this box for sometime and have finally managed to loose Windows Vista completely from the scheme.
My current issue is what RAID level to use for the system. My test install so-far has been RAID10,f2 with all six drives, giving 2250GB of usable space. I then used LVM to divide the space up into /,/usr,/opt,/var,/home with /boot on a separate ~128MB RAID1 at the top of each drive. This setup seems to allow about ~2014GB of free space on /home after the formatting and with all other volumes. My concerns are that this will not be enough space to accomodate all of the GIS data that I will likely accrue other the next year and also that I'm unsure how failure-proof the RAID10,f2 configuration really is.
I'm considering RAID6, as I know that would give me ~750GB more space, but at a cost of performance. I also know that RAID6 will take upto 2 drive failures. So, can the Linux implementation of RAID10 with layout f2 handle more than one drive failure? The idea of two far-copies seems to imply that only one failed drive could be sustained in this configuration.
So, again, my question is simply will software RAID6 create massive performance penalties on my 2xQuad (Xeon 2.8Ghz) workstation w/ 8GB FB-DIMM? I obviously need this system to perform well, but also need large storage capacity and fault-tolerance that can optimally handle multiple drive failures. I cannot add additional drives because of several reasons: (1) I'm broke, currently; (2) the onboard SATA controller is full; (3) the 1000W PS doesn't have anymore power connectors; and (4) the case would need additional HD brackets added to handle more than 6 HDDs.
Anyone have experience with RAID6 under Linux? I've seen several benchmarks showing major write performance issues, but those tend to be subjective and not show real world results.
Offline
So, a quick update.
I've decided to make several new arrays out of this configuration and then benchmark them. I was, however, reading that parity RAID is almost always going to encounter unrecoverable read errors during rebuilding with large drives. Is this true? Will my 750GB drives lead me to unrecoverable errors and not being able to rebuild the array?
Offline
So, another update.
I've decided to keep RAID10, add a PCI-E SATA controller with many ports and drop in additional discs as the needs grow. Then, I'll have the best of both worlds. Yes, it'll cost money that I don't have, but in the long run it's better to not dine out for a few weeks and not have to worry about the loss of data due to parity quirks. At the same time, I'll be getting better performance and that's always a plus on a machine that does heavy number crunching.
Offline