[comp.dcom.lans] Terminal Servers?

matt@iquery.UUCP (Matt Reedy) (12/08/88)

We have about 15 *nix boxes running a variety of flavors (Sys V/BSD/Xenix).
We'd like to have a single terminal in each programmer's office allowing
connections to any one of these machines (without having to run 15 cables from
each office).  I've vaguely heard of something called "terminal servers".  Is 
this what I'm looking for?  Anyone have any info on any brand in particular?

TIA.

matt
-- 
Matthew Reedy                 UUCP: {harvard!adelie,gatech!petro}!iquery!matt
Programmed Intelligence Corp.
400 N Loop 1604 E, Suite 330
San Antonio, TX  78232        (512) 490 6684

venters@argos.uucp (Jeff Venters) (12/09/88)

In article <147@iquery.UUCP> matt@iquery.UUCP (Matt Reedy) writes:
>We have about 15 *nix boxes running a variety of flavors (Sys V/BSD/Xenix).
>We'd like to have a single terminal in each programmer's office allowing
>connections to any one of these machines (without having to run 15 cables from
>each office).  I've vaguely heard of something called "terminal servers".  Is 
>this what I'm looking for?  Anyone have any info on any brand in particular?
>
>TIA.
>
>matt

There are several approaches you can take.  Here at CONVEX, we use both a
switch type and ethernet terminal servers.  The switch type requires more
cables since you must run one cable for every terminal you intend to support
(in all likelyhood it would be located in you computer room).  You then run
N lines to each host you want to serve.  With the ethernet terminal servers
you only need run one ethernet connection from each host.  The terminal
servers can be placed closer to the actual terminals (i.e., less wire).  As
for the user interface whenever you are not connected, if you type any key,
a menu of available hosts is presented.  These servers use "rlogin" to login
to the host.  I have been told by our operations folks that this approach
has a much smaller incremental cost per terminal than the switch type
server.  The ethernet terminal servers we use are made by CISCO (sp?).  The
switch type is made by MICOM.  I will say one good thing for the MICOM
server, it has been very reliable.  I can only recall it going down once or
twice in the last 2 years.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Jeff Venters, Convex Computer Corporation, Richardson, Texas	    
Internet: venters@convex.COM  |  UUCP: {uiucuxc,sun,rice}!convex!venters
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (12/10/88)

In article <147@iquery.UUCP> matt@iquery.UUCP (Matt Reedy) writes:
>We have about 15 *nix boxes running a variety of flavors (Sys V/BSD/Xenix).
>We'd like to have a single terminal in each programmer's office allowing
>connections to any one of these machines (without having to run 15 cables from
>each office).  I've vaguely heard of something called "terminal servers".  Is 
>this what I'm looking for?  Anyone have any info on any brand in particular?
>
	It depends.  Do you have serial ports on these machines and no
Ethernet?  If so, you need a port selector or data pbx.

	Want to yank the serial ports off and use Ethernet?  Then get
terminal servers.

	Otherwise, just hardwire everyone to a machine and use the
existing unix utilities to connect to other machines.

	Option three is cheapest, followed by option one.

DeadHead@cup.portal.com (Bruce M Ong) (12/10/88)

>We have about 15 *nix boxes running a variety of flavors (Sys V/BSD/Xenix).
>We'd like to have a single terminal in each programmer's office allowing
>connections to any one of these machines (without having to run 15 cables from
>each office).  I've vaguely heard of something called "terminal servers".  Is 
>this what I'm looking for?  Anyone have any info on any brand in particular?
>
>TIA.
>
>matt
>-- 
>Matthew Reedy                 UUCP: {harvard!adelie,gatech!petro}!iquery!matt
>Programmed Intelligence Corp.
>400 N Loop 1604 E, Suite 330
>San Antonio, TX  78232        (512) 490 6684

I am not sure about the exact details; but you would need to install ethernet
first; and you'll need terminal servers which support TCP/IP. Check with
Bridge Communications Inc (Now under 3COM), and ask about their CS/1 product
line. The phone is 1-800-ASK-3COM I think.

deadhead@cup.portal.com

mccalpin@loligo.fsu.edu (John McCalpin) (12/10/88)

At FSU, we have had good luck with the Annex terminal servers from
Encore.  They have 16 serial ports (non-standard connectors, though),
and one parallel printer port.  Our box is set up for a thick-wire
ethernet transceiver cable.
We have had no hardware failures in the 2 years that we have had the
box.  It does re-boot itself (by grabbing its software from a local Sun)
about every two to four weeks -- probably due to a soft parity error
in the 1 MB local RAM.  The software is pretty good, except that the
parallel printer driver does not seem too reliable.
Although all the ports are rated at 19,200 baud, the throughput
on the higher-numbered ports starts to degrade if more than
about 6 users are active at 9600 baud....  This is not noticeable
interactively, but kermit file transfers slow down a lot....

John D. McCalpin
mccalpin@masig1.ocean.fsu.edu

pavlov@hscfvax.harvard.edu (G.Pavlov) (12/12/88)

In article <12380@cup.portal.com>, DeadHead@cup.portal.com (Bruce M Ong) writes:
> ...... and you'll need terminal servers which support TCP/IP. Check with
> Bridge Communications Inc (Now under 3COM), and ask about their CS/1 product
> line. The phone is 1-800-ASK-3COM I think.

  I won't pretend to have any real experience with these.  But the overall
  sense of the net, as I interpret it, seems to be that "TCP/IP" and "Bridge"
  don't mix too well.  Encore and Cisco have gotten much more favorable res-
  ponse.

   greg pavlov, fstrf, amherst, ny

burzio@mmlai.UUCP (Tony Burzio) (12/13/88)

In article <689@hscfvax.harvard.edu>, pavlov@hscfvax.harvard.edu (G.Pavlov) writes:
>   I won't pretend to have any real experience with these.  But the overall
>   sense of the net, as I interpret it, seems to be that "TCP/IP" and "Bridge"
>   don't mix too well.  Encore and Cisco have gotten much more favorable res-
>   ponse.

We have a few Bridge servers and they work just fine through TCP.  I like
the fact that they boot themselves instead of waiting for a host to load
software (the base unit has a floppy drive with the programs in it).  The
only problem we have had is that ULTRIX kind of supports 8 bit mode, and
other UNIXs don't.  Leave everything 7 bits and you won't have problems.
We have experienced no hardware problems in 4 years (knock knock :-)

******************************************************************************
Tony Burzio          * To err is human...  To catch you at it requires a
Martin Marietta Labs * computer...  To fix it requires an accountant...
******************************************************************************

wilson@nova.laic.uucp (Robin Wilson) (12/14/88)

I was present at a presentation of the XYPLEX MAXserver line today.
They claim to support both TCP and LAT from the same board (and
individual serial connection).  Also, they had a pretty good layout for
management of the terminal servers (at least for my environment).  

My problem is, that I don't want to believe their presentation (I tend
to be pessimistic about a manufacturer's claims) so I need some other
input.  How 'bout it; anybody heard any horror stories about XYPLEX?

R.D. Wilson "My views -- hands off LMSC!"

jqj@oregon.uoregon.edu (J Q Johnson) (12/14/88)

In article <Dec.14.10.40.26.1988.112@ron.rutgers.edu>, ron@ron.rutgers.edu
(Ron Natalie) writes:
> Both Bridge and Cisco terminal servers will boot either from themselves
> (Bridge uses a floppy, cisco uses roms (look ma! no moving parts) or over
> the network.
It should be noted that Bridge/3Com terminal servers can boot themselves
over the network only from a special-purpose boot node (NCS) running XNS
(even if you are using a TCP/IP or OSI version of the Bridge t.s.). 
Furthermore, the boot server must be on the SAME XNS NETWORK, normally
the same Ethernet, as the terminal server.  In contrast, Cisco and Annex
terminal servers can boot themselves from any network node that supports
tftp on tcp/ip, even across tcp/ip gateways. 

This is a very substantial restriction on the Bridge servers in a typical
complex internet.

ron@ron.rutgers.edu (Ron Natalie) (12/14/88)

> We have a few Bridge servers and they work just fine through TCP.  I like
> the fact that they boot themselves instead of waiting for a host to load
> software (the base unit has a floppy drive with the programs in it).

Both Bridge and Cisco terminal servers will boot either from themselves
(Bridge uses a floppy, cisco uses roms (look ma! no moving parts) or over
the network.

-Ron

prc@maxim.ERBE.SE (Robert Claeson) (12/19/88)

ron@ron.rutgers.edu (Ron Natalie) writes:

> > We have a few Bridge servers and they work just fine through TCP.  I like
> > the fact that they boot themselves instead of waiting for a host to load
> > software (the base unit has a floppy drive with the programs in it).

> Both Bridge and Cisco terminal servers will boot either from themselves
> (Bridge uses a floppy, cisco uses roms (look ma! no moving parts) or over
> the network.

The Annex terminal servers will boot over the network from either a computer
or from another Annex.
-- 
Robert Claeson, ERBE DATA AB, P.O. Box 77, S-175 22 Jarfalla, Sweden
"No problems." -- Alf
Tel: +46 758-202 50  EUnet:    rclaeson@ERBE.SE  uucp:   uunet!erbe.se!rclaeson
Fax: +46 758-197 20  Internet: rclaeson@ERBE.SE  BITNET: rclaeson@ERBE.SE

phil@wubios.wustl.edu (J. Philip Miller) (12/20/88)

In article <442@maxim.ERBE.SE> prc@maxim.ERBE.SE (Robert Claeson) writes:
rutgers.edu (Ron Natalie) writes:
>> Both Bridge and Cisco terminal servers will boot either from themselves
>> (Bridge uses a floppy, cisco uses roms (look ma! no moving parts) or over
>> the network.
>
>The Annex terminal servers will boot over the network from either a computer
>or from another Annex.

and what about the security implications when these devices reboot over the
network?  What keeps some charlatan from providing an "improved" set of
definitions which destroys any secuity constraints which you have carefully
provided?  Forcing a device to reboot is frequently the easiest threat to
execute.

burzio@mmlai.UUCP (Tony Burzio) (12/20/88)

In article <280@wubios.wustl.edu>, phil@wubios.wustl.edu (J. Philip Miller) writes:
> and what about the security implications when these devices reboot over the
> network?  What keeps some charlatan from providing an "improved" set of
> definitions which destroys any secuity constraints which you have carefully
> provided?  Forcing a device to reboot is frequently the easiest threat to
> execute.

There is no such thing as security and networks.  There are enough
holes in UNIX that you should not worry about this little one...

*********************************************************************
Tony Burzio               * TriDECaphobia:  The fear of being burned
Martin Marietta Labs      * by buying DEC junk a fourth time...
mmlai!burzio@uunet.uu.net *
*********************************************************************

scarter@caip.rutgers.edu (Stephen M. Carter) (12/21/88)

>> Both Bridge and Cisco terminal servers will boot either from themselves
>> (Bridge uses a floppy, cisco uses roms (look ma! no moving parts) or over
>> the network.

>and what about the security implications when these devices reboot over the
>network? [....]

Almost anything can be broken into.  Given my sad all-night experiences
of updating a handload of Bridge CS-100 floppies, plus the fact that a
terminal server is not very high on a hacker's hit list, I'll prefer the
network load (from my choice of unix boxes too, thank you--not a special,
add-on, expensive boot device, aka server-server).

Two more points:
1) In Cisco's product, one could purchase the optional NV memory board for
the configuration information and use the prom to boot.  This still leaves
the option of a controlled mass configuration update via the network and
leave the network access off at other times for the security minded people.

2) In Bridge's case, you are mistaken if you think you have a secure system
by having the floppy.  A rather trivial 20 line program can get you Global
access on any Bridge on the network (even through gateways).

Stephen Carter
CAIP Center, Rutgers University

prc@maxim.ERBE.SE (Robert Claeson) (12/22/88)

In article <280@wubios.wustl.edu>, phil@wubios.wustl.edu (J. Philip Miller) writes:
> In article <442@maxim.ERBE.SE> prc@maxim.ERBE.SE (Robert Claeson) writes:

> >The Annex terminal servers will boot over the network from either a computer
> >or from another Annex.

> and what about the security implications when these devices reboot over the
> network?  What keeps some charlatan from providing an "improved" set of
> definitions which destroys any secuity constraints which you have carefully
> provided?  Forcing a device to reboot is frequently the easiest threat to
> execute.

Well... You can

- tell the box to boot from a specific server by default.

- set a password so you can't reconfigure it without knowing the
  password, neither from the network nor from the Annex.

- disable any commands you don't want someone to see.

- enable the ACP (Access Control Protocol).

- enable the logging facilities.

- etc, etc.
-- 
Robert Claeson, ERBE DATA AB, P.O. Box 77, S-175 22 Jarfalla, Sweden
"No problems." -- Alf
Tel: +46 758-202 50  EUnet:    rclaeson@ERBE.SE  uucp:   uunet!erbe.se!rclaeson
Fax: +46 758-197 20  Internet: rclaeson@ERBE.SE  BITNET: rclaeson@ERBE.SE

prc@maxim.ERBE.SE (Robert Claeson) (12/22/88)

In article <444@maxim.ERBE.SE>, prc@maxim.ERBE.SE (Robert Claeson) writes:
> In article <280@wubios.wustl.edu>, phil@wubios.wustl.edu (J. Philip Miller) writes:

> Well... You can
> 
> - tell the box to boot from a specific server by default.
> 
> - set a password so you can't reconfigure it without knowing the
>   password, neither from the network nor from the Annex.

Oh, forgot to mention that the configuration is stored in an EEPROM
in the Annex.

-- 
Robert Claeson, ERBE DATA AB, P.O. Box 77, S-175 22 Jarfalla, Sweden
"No problems." -- Alf
Tel: +46 758-202 50  EUnet:    rclaeson@ERBE.SE  uucp:   uunet!erbe.se!rclaeson
Fax: +46 758-197 20  Internet: rclaeson@ERBE.SE  BITNET: rclaeson@ERBE.SE

sob@watson.bcm.tmc.edu (Stan Barber) (12/28/88)

In article <109@oregon.uoregon.edu> jqj@oregon.uoregon.edu (J Q Johnson) writes:
>It should be noted that Bridge/3Com terminal servers can boot themselves
>over the network only from a special-purpose boot node (NCS) running XNS
>(even if you are using a TCP/IP or OSI version of the Bridge t.s.). 
This is incorrect. The NCS does NOT run XNS. I have an NCS/AT that runs
TCP/IP just fine, thank you.
>Furthermore, the boot server must be on the SAME XNS NETWORK, normally
>the same Ethernet, as the terminal server.  
The must be on the same ETHERNET, not the same XNS network. If you have
a large network that uses bridges or bridging routers, it all works.
>This is a very substantial restriction on the Bridge servers in a typical
>complex internet.
I agree that the 3Com/Bridge boot system is not the best, but let's try to
argue from accurate information.




Stan           internet: sob@bcm.tmc.edu         Baylor College of Medicine
Olan           uucp: {rice,killer,hoptoad}!academ!sob
Barber         Opinions expressed are only mine.

jqj@oregon.uoregon.edu (J Q Johnson) (12/28/88)

In <1357@gazette.bcm.tmc.edu>, sob@watson.bcm.tmc.edu (Stan Barber) replies:
>>It should be noted that Bridge/3Com terminal servers can boot themselves
>>over the network only from a special-purpose boot node (NCS) running XNS
>>(even if you are using a TCP/IP or OSI version of the Bridge t.s.). 
> This is incorrect. The NCS does NOT run XNS. I have an NCS/AT that runs
> TCP/IP just fine, thank you.
Interesting.  My CS200s (software v 20000) use UDP/IP for everything *except*
booting.  For booting, they send out XNS IDP packets with network number 0.
 What version are you running?

>>Furthermore, the boot server must be on the SAME XNS NETWORK, normally
>>the same Ethernet, as the terminal server.  
> The must be on the same ETHERNET, not the same XNS network. If you have
> a large network that uses bridges or bridging routers, it all works.
I stand by my phrasing.  They must be on the same XNS network as defined
by a this-net (net 0) IDP broadcast.  Agreed that this network may contain
several segments, and hence need not be "the same ETHERNET".  The important
thing is that it may not be an XNS internet -- a Hybridge is OK, but not
an XNS router.
> I agree that the 3Com/Bridge boot system is not the best, but let's try to
> argue from accurate information.
Agreed.

Other problems with the 3Com/Bridge design: the bootstrap process (which
uses XNS) does not appear to be able to cope well with lossy networks;
the "remote" network management protocol (which uses UDP/IP) has
effectively no security except obscurity of the protocol.

sob@watson.bcm.tmc.edu (Stan Barber) (12/29/88)

After some private mail on this issue, I am rereplying:
]In article <109@oregon.uoregon.edu> jqj@oregon.uoregon.edu (J Q Johnson) writes:
]>It should be noted that Bridge/3Com terminal servers can boot themselves
]>over the network only from a special-purpose boot node (NCS) running XNS
]>(even if you are using a TCP/IP or OSI version of the Bridge t.s.). 
]This is incorrect. The NCS does NOT run XNS. I have an NCS/AT that runs
]TCP/IP just fine, thank you.
The NCS/AT runs TCP/IP most of the time. XNS is used to boot only.
]>Furthermore, the boot server must be on the SAME XNS NETWORK, normally
]>the same Ethernet, as the terminal server.  
]The must be on the same ETHERNET, not the same XNS network. If you have
]a large network that uses bridges or bridging routers, it all works.
XNS is defined but not supported consistantly by all vendors. 3com XNS
and Bridge XNS and Xerox XNS do not interoperate well. It is the case
that Proteon's and cisco's XNS support may not support all these 
different forms of the protocol. (I know that Proteon only claims to
support the XNS used by Novell.) So, we are both probably correct.
>This is a very substantial restriction on the Bridge servers in a typical
>complex internet.
]I agree that the 3Com/Bridge boot system is not the best, but let's try to
]argue from accurate information.

Anyone want to contribute to the accurate information pool, please do.







Stan           internet: sob@bcm.tmc.edu         Baylor College of Medicine
Olan           uucp: {rice,killer,hoptoad}!academ!sob
Barber         Opinions expressed are only mine.