[comp.dcom.lans] Twisted Pair Ethernet Concentrators

matthews@uswat.uswest.com (John Matthews) (09/07/90)

A group of us at US West Advanced Technologies are currently evaluating
twisted pair ethernet concentrators.  What we would like to know is
what vendors do you know of that have the following features?  We are
also looking for people that are familiar with Chipcom, Fibermux and
ODS that would be willing to share their experiences.  These features
that I am asking about are not absolutely mandatory, I just want to
know what vendors have implemented what features.  Alot of vendors
we have been talking to claim to support a subset of the features
mentioned below.
			Thanks in advance,
				John Matthews
				matthews@uswat.uswest.com



Appletalk -      It would be nice to have reliable Star Controller
		 functionality incorporated within the same box, but
		 this is mostly wishful thinking.

Stability -      Ability to withstand shorts in wiring, power supply burnouts,
	   	 etc.  Overall, we need a very reliable concentrator.

Multiple Segment Capability - The ability for one chasis to support three
		 or four different logical ethernet segments.  These
		 segments will be connected to each other via a router.

Node Detect - Whenever a concentrator sees a source MAC address
		 of an ethernet device transmitting that does not
		 match the address registered with that port, send an
		 alert stating what that ethernet address is.  We would
		 like the option of also having this port disabled
		 immediately in such a case.  A threshold would be
		 nice with this feature so that we could say after
		 seeing X packets with a non-registered MAC address
		 go ahead and disable the port.  This would be useful
		 in the case where a network error could look like a 
		 bad ethernet address.

Node UP Integrity - Whenever the concentrator first sees the line
		 integrity signal go on on a port, start a timer to wait
		 and see if the attached node transmits within a
		 certain amount of time.  If it doesn't, send an
		 alert to the network management console and also
		 give us the option of disabling the port.

Threshold Shutoff/Warning - We would like to be able to receive
		 alerts whenever given hosts transmit at a rate
		 greater than a given threshold.  An extension to
		 this feature that would of great importance would
		 be another threshold with broadcasts.  If a given
		 node starts broadcasting higher than a given
		 threshold, send a warning and disable the port.

Automatic Port Disable - The ability to disable a given
		concentrator port whenever the line integrity
		signal has disappeared for more than a given amount
		of time.  This should have a per port configurable
		threshold.  This would be useful in the case that
		users unplug or remove equipment without notifying
		network operations.  Equipment is often moved and
		taps to the network are left wide open.

Port Idle Time - The ability to monitor how long each port on
		the concentrator has been idle.  In other words,
		how long has it been since data was last transmitted
		from the node connected to the twisted pair transceiver.
		
Generic Access - The ability to telnet into the box and get
		quick status information.  We are often in remote
		parts of the building or other networked sites and
		it would be nice to be able to access this equipment
		easily through the network.  Async access is an
		alternative, but not a good one.

Port Level Analysis - Exactly what statistics can your equipment provide
		for each port?

Software Upgrade Strategies - We would like to see a popular
		boot protocol such as tftp being implemented.
		Such a protocol must work through our IP routers.

SNMP Manageability -
		Must support SNMP queries and changes.
		Although some of the features mentioned above could
		be implemented with an SNMP monitor by polling, we
		would like alot of these features to work independent
		of an SNMP management station.  If the segment (or any
		hardware in between) connecting the SNMP monitor ever
		went out, our security and manageablity would be cut-off.

kwe@buit13.bu.edu (Kent England) (09/07/90)

In article <10898@uswat.UUCP> 
matthews@uswat.uswest.com (John Matthews) writes:
>
>A group of us at US West Advanced Technologies are currently evaluating
>twisted pair ethernet concentrators.  What we would like to know is
>what vendors do you know of that have the following features?  ...

	I estimate that the computational power required to fulfill
your wish list would be roughly that of a midrange mini or LAN server.

	LAN hardware should be a little simpler than what you are
thinking, in my opinion.  Keep the complexity in the end systems and
intermediate systems.  That idea is the essence of the LAN paradigm
over earlier models of networking.

	Why don't you substitute RS-232 ports for the Ethernet and
LocalTalk ports and call it timesharing?  :-)

	--Kent