NanoKVM: Affordable, Multifunctional, Nano RISC-V IP-KVM
github.com
github.com
It's probably 2-4 weekends worth of work for someone who already worked with all of this in the past to combine it together and add a basic web interface. (Assumming they will use sophgo's Linux kernel branch and will not try to mainline kernel support, first)
Well that sucks then. I hoped it's in similar state like Rockchip, where the required things already work over V4L2 API (in both BSP and upstream).
But that just means that Linux upstreaming will have to come first. I personally don't see why anyone would want to waste time with some random vendor's media API and its quirks when it will get replaced by v4l2 in the future.
To me 20/80 also means that I need 20% of the features of a serious product and prototype is often enough, as long as it's reliable. :)
There are a few around, but your phrasing makes it sound as if closed source/binary isn't the default for these kinds of devices
So would I. I've used several of the standard ones for 20 years, and the closed nature universally makes it suck.
They always require some ancient java install or a license of one form or another, I have one that needs IE 6 with ActiveX! (AMI MegaRAK, standalone unit not built into a motherboard) And of course that thing only has ancient ssl and ssh and neither can be updated. Sure it's old, but there's nothing about it's job that changed or got harder requiring new hardware. I don't use it any more only because of things that are the result of it's closed nature. I actually really like the hardware. If I could replace the firmware I would still use it. Nothing about the job it does has changed, and so it doesn't matter how old the hardware is.
When I discovered PiKVM a few years ago before he was selling a product yet, I built one and it replaced a $600 Lantronix immediately.
This was a personal one not one work paid for, I already owned a spiderlinx, actually 2, one for vga & ps2, and one for dvi & usb, I already own these and I had paid the 500-600 retail for them, and a PiKVM built myself from a pi and a capture board from aliexpress and hacked into a generic aluminum box style pi case that I cut and drilled, is better. It's worth more in that it provides more functionality and flexibility.
Then I donated $500 to him, because that was still less than the cost of an equivalent, and now I get to have as many as I want for just the cost of the hardware, which I am free to scrounge up out of anything if I have to, and I'd wanted something like that for years and here it finally was. It was a real pain point for years and I was grateful and want to reward the things I want to exist.
And now that he sells a product with custom pcb, the entire kit including the pi is still about 1/2 of a spiderlinx.
You're right that it's normal that none of them disclose their firmware, or let you replace it, and it absolutely sucks all day every day every minute you have to use any of those normal ones.
For work before cloud times when we had all colo, what actually worked the best was serial consoles. You set up a little serial port server which you can ssh in to (and keep updated with current ssh) and it provides a serial connection to each machine in the rack. No java, no atvivex, no special browser, no licence manager, no software or platform requirements at all. Anything that can ssh can do a bare metal reinstall. And it works fine over a crappy cell connection.
But if you're forced to deal with a Windows box, or something else with only a screen & keyboard like a security camera console or a mac or something, or just any random non-server pc without serial console support in it's bios, if you do need a kvm, pikvm or equivalent is the only way to go.
https://github.com/pikvm/pikvm
tinypilot is open source (but they kind of restrict "pro" binary availability)
You can make a pikvm for almost free if you want, because it doesn't care what hardware you use. You probably either already own or at least have access to everything necessary for free. That fact is 100x more valuable even aside from the price, simply the fact that you can source the necessary hardware from anywhere at any time on zero notice. When things go wrong, you are far more likely to be able to lash up a solution immediately if all you need is any kind of computer and any kind of capture device, and you can even tell a remote customer what to buy at their local Walmart if you had to.
But even the full finished productized package with custom hat pcb and case is still 1/2 of the equivalent single port Lantronix, and is more convenient and more useful.
And even the proprietary code in something like a Lantronix which is a long standing industry name with accountability and a reputation to protect, is still a 100% different proposition than the proprietary code from an unknown nobody.
It would be crazy, today, completely grossly negligent irresponsible, to use such a device in a high consequense role as remote server administration.
An ip-kvm is fundamntally literally a man in the middle and a keylogger, which you use to access all the most critical things that everthing else happens on. No big.
Edit: and it looks like the image for this is not open source.
I'm pretty sure all my 10G equipment goes down to 100M; 10M support is often dropped or at least not documented as available though. I'd be surprised if many switches dropped 100M support, because most tvs and set top boxes are 100M, and there's a huge amount of other still useful devices that didn't get gigE. OTOH, i think I've only got three devices that are 10M only: my gate controller, my dreamcast LAN adapter, and a really old dev sbc (rabbit something).
I guess that's because consumers want HDMI capture cards so that they can record their PS5 gaming session, but not DP capture cards?
I'd much rather be able to attach something like this, have it spin up its own wifi or ethernet network (that understandably wouldn't have internet access) and then connect from my laptop. The other solution I was considering is a portable monitor (one of those hand-held tablet-sized ones) which absolutely would work without a network but the NanoKVM seems less wasteful in terms of future e-waste generated.
Yes, it is.
Restart on power loss in the bios is usually the right way if the machine is a server that you know you always want it to be on or trying.
It only gets complicated when you have say, a full backup clone of that machine that you actually want to be a cold spare not a hot spare. So it exists, is plugged in, but turned off, and you want it to stay off unless you explicitly turn it on because it would conflict with the primary, and yet want to be able to do that remotely or even automatically but by some more complex logic than merely the fact that power went from off to on.
That's not super common, but it would take something weird like that to not want simple bios always-on.
Usually it should be easy enough to have a machine always try to turn on, and decide for itself that it should shut back down.
My use-case is, I have some computers I use for a home lab. If I'm outside of the house and something happens to those computers that requires manual intervention, there is nothing I can do. With this, I can power the computer off and on (via a smart plug, not the KVM), then see and interact with the entirety of the boot process, including accessing BIOS, select OS on GRUB, etc. As long as the KVM has network connection, it serves as a remote interface to the phisical computer.
- Higher encoded framerate
- Loading bootable ISOs directly via the IP KVM isn't a per-BMC model dance (if the BMC does it at all, without begging for you to hack the BMC license)
- Similarly, loading drivers doesn't require making a compatible image file or finding another USB
- The web interface doesn't have any legacy crap (java/.net/activex) left in the menus
- I can cold boot a server and not lose my session
- You can paste large blocks of text and it will type it out
- You can text extract large blocks of text from the framebuffer
- You can just load standard linux packages for whatever you need in a pinch (on the PiKVM at least)
- You can actually update and secure this thing vs hoping your BMC either receives real updates or someone hacks it so you can load an open source one instead.
I'd say the main downside is the ATX power control isn't something you just "plop in". You either wire it up and use the device with the same server over long periods of time or you don't use the ATX power control.
This Youtube really helped: https://www.youtube.com/watch?v=mhq0bsWJEOw
And there's more here: https://www.meshcommander.com/meshcommander/tutorials
It can also be really handy (with storage support) to quickly install a set of machines in a rack - either have them boot off the internal USB storage or go into the BIOS to monitor PXE boot progress and handle any input required.
* Evading corporate IT: I do not recommend doing this.
* Normal[0] remote access to a machine: Yeah, running a remote desktop in software on the machine[1] is almost always going to be easier, cheaper, more performant, and more flexible than anything you tack on externally.
[0] Where "normal" remote access means it doesn't need to work when the machine is kernel panicked or in the firmware setup screens.
[1] The machine itself can be whatever you want. It can be an always-on desktop, but it can also be a laptop that sits in the corner and that only boots up when you poke it with a Wake-on-LAN packet, or a VM on an ESXi cluster, or a (carefully secured) Hetzner Cloud VM. That part is dealer's choice depending on your needs.