[sclug] kvm switch (> 2 port, electronic)

[Matt Dainty]

On 8 Jan 2007, at 18:27, ed wrote:

> totally irrelevant to the question, but i've had recent experience of
> the Dell 4161DS. to begin with i set up some ssh tunnels to the
> remote location and did all the magic so that i could access it, but
> underneath it's really quite a pain:

I've used something similar quite a lot, the 2161DS. TBH, it's just a  
rebadged Avocent KVM.

> 1) cannot connect to more than 4 remote boxes at once, despite it  
> being
>    labelled as suitable for 128 devices
>
> 2) the limit of 4 is total number of clients, so if bob is using the
>    kvm at the same time, we can only use a total of 4 remote computers

There's a similar limit with this model, but I've rarely hit it, the  
KVM tends to be used as a last resort, so if it's only if the machine  
appears to be dead on the network.

> 3) the mouse pointer sucks, some alternative pointing device would be
>    better, like a puck from the old days.
>
> 4) it is not always possible to use devices that are on the same 'pem'
>    module. the KVM data is on UTP cable, and we know how much outside
>    interference effects video... so the 'pem' devices are little
>    repeaters for the signal. if bob is using a computer on the same
>    pem, it's not always possible for someone else to work.

No, I don't think you _can_ use more than one device on the same PEM  
module, the concurrent client count is based on the physical ports on  
the KVM, so for example you can have 2 concurrent users on any two  
physical KVM ports, so if the two devices are hung from the same PEM  
module, no dice.

The other problem is that the way the KVM works is by rasterising the  
picture to send it down the wire, and as you've noted, you can get  
interference on the wire, so if the screen starts flickering, the KVM  
sends those flickers down the wire to you which really can knacker  
your bandwidth. I've seen it kill a 2Mb/s leased line.

> 5) the price tag is $5,000; http://s5h.net/u?f25
>
> there are some nice things, like the software is java based, so it can
> be used on most terminals. but the above things are just so damn ugly
> that it promises much more than it delivers.

It's Java based yes, but sadly the actual KVM innards are either a  
Windows DLL or a Linux/i386 .so library so it doesn't work on a PPC  
Mac, for example :-( Just shows that Java != Portable everywhere.

Having said that, the little SIP modules for each server makes  
cabling up a rack quite tidy, there's no bulky KVM cables, just a  
piece of Cat5 to route wherever.

> it didn't come from my pocket, and for the location that i'm  
> working in
> right now, a kvm is necessary, but these little points are so painful.

Yeah, I don't think they're great either, but all the servers are  
Dell so if they also do a KVM, it goes on the order as it's easier  
than having multiple suppliers/orders. Sadly Dell seem to rebadge  
someone elses hardware to just say "Yes, we have a KVM-over-IP  
product", they don't appear to support or maintain it terribly well.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: This is a digitally signed message part
Url : http://lists.tmdg.co.uk/pipermail/sclug/attachments/20070109/8539e314/PGP.bin