USB/IP Project • Hacker News Board

Re: USB/IP Project

#21

Post by orik » Tue, Oct 13, 2015, 12:20 AM UTC

kenz0r wrote: post #18

pg_is_a_butt wrote: post #7

seems the network latency would break the realtime expectations of high data rate USB 3.0 hardware on the USB bus. and using a still image from a webcam seems like the worst use case... there are much better protocols for serving video streams.

One of the more compelling use-cases for USB over IP is connecting proprietary licensing dongles to virtual machines.

can't you use vt-d and have it passthrough?

Re: USB/IP Project

#22

Post by pudquick » Tue, Oct 13, 2015, 12:20 AM UTC

I would love to see this develop into a widely adopted standard. Apple uses a similar technology with their Airport Express devices for printer sharing, always been annoyed it was so restrictive in being able to connect to it.

Re: USB/IP Project

#23

Post by kenz0r » Tue, Oct 13, 2015, 12:30 AM UTC

orik wrote: post #21

kenz0r wrote: post #18

Earlier quoted context omitted.

One of the more compelling use-cases for USB over IP is connecting proprietary licensing dongles to virtual machines.

can't you use vt-d and have it passthrough?

That still ties it to a physical VM host doesn't it - what happens if you want to host the VM in a cluster and have it migrate?

Re: USB/IP Project

#24

Post by jjoonathan » Tue, Oct 13, 2015, 12:31 AM UTC

ChuckMcM wrote: post #19

I've got a couple of the Chinese made "Remote USB" servers which are interesting but not as compatible as I would like. What I find really amusing is that back when USB was proposed I argued that rather than create an entirely new infrastructure why not just use Ethernet packets. There are additional protocol types available in the bit encoding and all sorts of infrastructure that already supported layer 2 switching.…

It's been funny to watch the various phy standards converge in feature-space (alas, not interop-space) over time. PCI got a packet architecture with addresses and routing and credit-based flow control, USB got a packet architecture with addresses and routing and credit-based flow control, SATA said "screw it, we'll just become PCI-e, because they have packets and routing and flow control," and thunderbolt basically is PCI-e...

Unfortunately I think Ethernet might be too fragmented to ever take the convergence to its logical conclusion. But who knows. If someone high on the food chain at Intel wakes up on the other side of the bed one day it could happen.

Re: USB/IP Project

#25

Post by DiabloD3 » Tue, Oct 13, 2015, 12:34 AM UTC

Can I have the reverse? IP over USB at USB 3.x speeds?

Re: USB/IP Project

#26

Post by roel_v » Tue, Oct 13, 2015, 12:41 AM UTC

LordHog wrote: post #20

roel_v wrote: post #10

Just buy one of these: http://www.amazon.com/Silex-SX-DS-4000U2-Print-Server-800MHz... . Works awesome, lets you buy the cheap non-workgroup version of pretty much everything (printer, scanner, label printer, ...) and still hook it up to a central server. I have a Fujitsu ScanSnap ix500 hooked up this way, which scans directly to Dropbox. Put the scanner anywhere where it's convenient to access, put paper in it, pres…

At ~ $150, that is an expensive solution, IMHO.

Depends what you compare it against, I guess. Seemed like a bargain to me.

Re: USB/IP Project

#27

Post by ausjke » Tue, Oct 13, 2015, 12:54 AM UTC

DiabloD3 wrote: post #25

Can I have the reverse? IP over USB at USB 3.x speeds?

yes on linux(usbnet) at least for a long while, not sure about usb3.x or windows support though.

https://en.wikipedia.org/wiki/Ethernet_over_USB

A quick google showed there are Gigabit NIC over USB3.0 on the market already(for a while): http://www.amazon.com/Network-Adapter-Anker-Ethernet-Support...

Re: USB/IP Project

#28

Post by SXX » Tue, Oct 13, 2015, 12:55 AM UTC

orik wrote: post #21

kenz0r wrote: post #18

Earlier quoted context omitted.

One of the more compelling use-cases for USB over IP is connecting proprietary licensing dongles to virtual machines.

can't you use vt-d and have it passthrough?

You don't even need VT-d for USB devices, only PCI one.

Re: USB/IP Project

#29

Post by DiabloD3 » Tue, Oct 13, 2015, 12:59 AM UTC

ausjke wrote: post #27

DiabloD3 wrote: post #25

Can I have the reverse? IP over USB at USB 3.x speeds?

yes on linux(usbnet) at least for a long while, not sure about usb3.x or windows support though. https://en.wikipedia.org/wiki/Ethernet_over_USB A quick google showed there are Gigabit NIC over USB3.0 on the market already(for a while): http://www.amazon.com/Network-Adapter-Anker-Ethernet-Support...

I don't want a USB NIC, though, I want just USB to USB with as little shit in between them as possible to be as cheap as possible. Maybe even allow straight DMA to DMA across the machines due to it being USB 3.x.

Re: USB/IP Project

#30

Post by simcop2387 » Tue, Oct 13, 2015, 1:09 AM UTC

jjoonathan wrote: post #24

ChuckMcM wrote: post #19

I've got a couple of the Chinese made "Remote USB" servers which are interesting but not as compatible as I would like. What I find really amusing is that back when USB was proposed I argued that rather than create an entirely new infrastructure why not just use Ethernet packets. There are additional protocol types available in the bit encoding and all sorts of infrastructure that already supported layer 2 switching.…

It's been funny to watch the various phy standards converge in feature-space (alas, not interop-space) over time. PCI got a packet architecture with addresses and routing and credit-based flow control, USB got a packet architecture with addresses and routing and credit-based flow control, SATA said "screw it, we'll just become PCI-e, because they have packets and routing and flow control," and thunderbolt basically i…

Ethernet's also got the fun problem of the length of the runs being hugely different than pcie. What I expect is some kind of optical interface because of that. Optical thunderbolt might be the way for that to happen.