Hi All,
I am interested in setting up a freeNAS box whose sole purpose will be to receive rdiff-backup data from another machine connected via gigabit ethernet in a separate building. Thus the data will arrive in a burst once every 24 hours, and will rarely be read back off of the freeNAS box. The bandwidth also sets a pretty tight bottleneck on read/write speed, so I am not interested in "optimum" ZFS setup if the benefit will be clobbered by the network bottleneck anyway.
After reading the manual, I have three remaining questions:
1) How important is the "operating system" drive? Is there any reason to prefer compact flash over just a USB flash drive? (Is getting CF support on a rackmount system even feasible?). Does the USB speed matter at all? Will I regret just buying a cheap low-profile usb stick and leaving it in the back of the sever?
2) Performance. It looks like optimal performance is achieved with 2^n+1 storage drives. Given the intermitent nature of my data access, is the performance penalty from using say, 6 drives in a raidZ arrangement significant? (5 drives worth of data, 1 of parity).
Additionally, the rule of thumb I'm seeing is 8GB of ram for up to 6TB of storage space, plus 1GB for every TB of storage beyond 6TB. How concerned should I be about that given my use case of occasional transactions over a single gigabit link?
3) I plan to run rdiff-backup over ssh, so I am not planning to set up NFS shares or samba or anything. How do permissions work in this case? Can I just set up a user called "rdiff" and one big dataset which that user can read/write?
Any tips/suggestions/feedback are much appreciated!
I am interested in setting up a freeNAS box whose sole purpose will be to receive rdiff-backup data from another machine connected via gigabit ethernet in a separate building. Thus the data will arrive in a burst once every 24 hours, and will rarely be read back off of the freeNAS box. The bandwidth also sets a pretty tight bottleneck on read/write speed, so I am not interested in "optimum" ZFS setup if the benefit will be clobbered by the network bottleneck anyway.
After reading the manual, I have three remaining questions:
1) How important is the "operating system" drive? Is there any reason to prefer compact flash over just a USB flash drive? (Is getting CF support on a rackmount system even feasible?). Does the USB speed matter at all? Will I regret just buying a cheap low-profile usb stick and leaving it in the back of the sever?
2) Performance. It looks like optimal performance is achieved with 2^n+1 storage drives. Given the intermitent nature of my data access, is the performance penalty from using say, 6 drives in a raidZ arrangement significant? (5 drives worth of data, 1 of parity).
Additionally, the rule of thumb I'm seeing is 8GB of ram for up to 6TB of storage space, plus 1GB for every TB of storage beyond 6TB. How concerned should I be about that given my use case of occasional transactions over a single gigabit link?
3) I plan to run rdiff-backup over ssh, so I am not planning to set up NFS shares or samba or anything. How do permissions work in this case? Can I just set up a user called "rdiff" and one big dataset which that user can read/write?
Any tips/suggestions/feedback are much appreciated!
