The 6351 gives you 1 MHz / N Channels (when N>1). So at 10 channels you would indeed be able to sample each at 100 kHz (keep in mind it's MUXed so there is an offset between channels). There is a USB variant of this as well.
You can't run the channels at unique sample rates per se... but you could configure a task with a scan order like this:
fast channels are {A, B, C, D}
slow channels are {1, 2, 3, 4}
A B C D 1 A B C D 2 A B C D 3 A B C D 4 ...
clocked @ 1 MHz
It would give 200 kHz for A, B, C, and D while giving 50 kHz for 1, 2, 3, and 4. It's kind of a niche use case but might help you out.
The 9220 would be an alternative to the 9205 if you prefer the C Series form-factor. It gives 100 kHz per channel with 16 channels of input. It is not multiplexed so does not have the same channel-to-channel delay issue. All of the channels sample together off of the same clock so they must also run at the same rate.
Unless you have a compelling reason to use Linux I'd suggest running Windows if possible as it has better support from NI. Looking at NI's support for Linux for both PCI/PXI and Portable devices, your best Linux option is probably the PCI / PCIe-6251 which will perform similarly to the 6351.
There is the RT option as well that you might consider (cRIO / PXI seems excessive to me for this application and the standalone cDAQ chassis are quite expensive). There's a pretty substantial software cost though.
Best Regards,
John Passiak
