An anonymous reader asks: "I've been looking for a cost effective (ie, cheap) way to remotely administer several servers running a variety of OS's, and would like to have a solution that would allow for monitoring of the bios on startup, etc (ie, not VNC). The most appealing solution is KVM over IP, which really just means a souped up KVM switch with something like VNC running on it, unfortunately all of the solutions I've been able to find are more expensive than I can justify spending. I've played around a bit with making my own Poor man's KVM over IP; I did this by purchasing a cheap (sub $50) VGA-to-NTSC convertor, then feeding it into a video card with NTSC input (the ATI All-In-Wonder Radion), and then by logging into a machine running Windows Terminal Services I'm able to watch the reboot process. Of course, this doesn't address the mouse/keyboard issue, and the quality isn't all that great. What I'm hoping is that someone else might have a suggestion on how to do this, preferably using Linux and the least hardware necessary. Does anyone have any suggestions or insights on ways to do this?" There are pre-existing solutions, but it seems they are all kind of pricey. Can any of you suggest cheap solutions (at or below $500USD) that could handle a farm of 5-10 machines?
"Here are the three approaches I found:
ViewProxy:
They make the most economical for administration of multiple machines (by one person). Their ProxyView device plugs into your KVM just like it was a monitor/mouse/keyboard, and then does all the packetizing magic. Price is about $6k from what I can tell.
eRIC:
These are the same guys who make the Rolf (Reboot on Lan), which is pretty cool. They make a card called Eric which replaces your normal video card with their card, which has a built in ethernet connection and allows remote control. The cheapest solution at about $700 but only would allow control of the machine it's installed in.
Avocent: I think the first to introduce the whole KVM over IP solution, they have KVM's with this sort of functionality integrated. Some of their products allow multiple users to multiple machine, which is a neat feature but not needed for my applications. Their units run from $4k on up."
Bandwidth Issues? (Score:2, Insightful)
Cheap alternative (Score:2, Insightful)
Check out Raritan [raritan.com]. They have a wide range of such products. Not sure about prices though.
'Ask Slashdot' has taught me something. (Score:5, Insightful)
The Right Way involves spending a little more money up front, but its benefits are manyfold: A proven solution, vendor support, reliability, stability, and various and sundry other good things.
The Slashdot Way involves duct tape, bailing wire, and, sometimes, a 386 running RedHat. Its generally insignificant up-front savings are offset by the countless hours of configuration, tuning, tweaking, prodding, poking, and general lackluster performance of the contraption in question.
You have chosen to go The Slashdot Route. I wish you luck as you set up your TV cards and serial ports. You will need as much luck as you can get, and an awful lot of patience.
- A.P.
Re:They sell them here... (Score:1, Insightful)
WHY DO PEOPLE DO THIS ??!!??
We had a KVM extender that used CAT5 to connect the remote box to the box in the server room.
Nice feature: If you pay attention to what you are doing, you can use the cable that you already pulled to connect these things.
Sucky feature: Sooner or later someone will connect the KVM connector box to the wrong port in the wall, and fry the box. (it probably won't do your router too much good either) (it didn't take us too long)
This is exactly what murphy was talking about when he came up with that law of his.
It doesn't seem fair, but any piece of equipment with a CAT5 cable on it that can't deal with being plugged into an IP router w/o blowing itself or the router should be destroyed immediatly.
Re:'Ask Slashdot' has taught me something. (Score:5, Insightful)
However, this doesn't make for "The Right Way". Hacking at something - figuring out how it works, seeing how you can do it better (or less expensively), and enjoying the process - is the source of solutions that Just Work.
"[G]eneral lackluster performance of the contraption in question" is the result of not understanding something enough to do it well. Many off-the-shelf solutions suck - Windows 98, anyone? So do many home-brew setups. The problems doesn't come from the nature of a rig, it comes from the effort and intelligence of the creator.
If you prefer to not think about things, and just have them work, fine - but don't disparage people who are interested in learning and improving with sweeping and inaccurate generalities about DYI projects. Most of your "proven solutions" only got to that point because enough people (or sometimes, one intelligent person) was willing to hack at something until they were satified.
End of rant.
Yes... but when he's done he'll have a clue... (Score:3, Insightful)
A few years down the road and most of us will want to hire the experimenter who has tried several different OSes, hacked out a wireless network out of a couple 2-meter transceivers, set up two 486 DX66 boxes as a dedicated VPN between his bedroom and his girlfriend's house, and wired up the girl's locker room with x10. Those are the guys who can think their way through a problem rather than hitting the catalogs looking for a million dollar solution.
Sneaker Net, are you fucking stupid! (Score:1, Insightful)
What did it cost? (Score:3, Insightful)
It worked well, but REALLY expensive for a 16 port version. Expensive to the tune of around $10k for the box, the auth server module, and 2-3 client licenses. I was most turned off by the fact that the server and client software were $old $eperately, since the software is useless without the hardware.
I read a USENET post (circa 11/2001) that said the devices were buggy and the vendor was an asshole about other platform clients and future development/changes.
I think digitized video and IP KVM connectivity is probably not a fluke and represents the "future" of KVM, but vendors will need to seriously get their shit together in terms of client access and pricing otherwise computer makers are just going to crush this product with their own built-in remote management. All our HP servers have built-in serial management that can do power on/off/reboot, environment management, and text/keyboard redirection; HP and Compaq both have boards that can do it natively over IP, the *only* thing missing is the ability to do transparent video redirection. When they do that, KVM will be obsoleted by a laptop running a redirection client.
Good, but wrong. (Score:3, Insightful)
No. Generally, in my experience, it's the source of solutions that Almost Work. Or solutions that Work Unless You Do This. Or solutions that Just Worked Last Week, What the Hell Did We Change That Broke It?
"[G]eneral lackluster performance of the contraption in question" is the result of not understanding something enough to do it well. Many off-the-shelf solutions suck - Windows 98, anyone? So do many home-brew setups. The problems doesn't come from the nature of a rig, it comes from the effort and intelligence of the creator.
No, the problem comes from the continual poor reinvention of the wheel on Ask Slashbots. In this situation, KVM-IP switches are the answer. Not a 486 with a bunch of TV cards in it. Not a rat's nest of cables. If this person worked for me and proposed this solution, I would have a hard time signing his checks from then on.
- A.p.