Cell Relay Archive

[crystal <crystal@postme.net>]

Data compressors, concentrators, multiplexors, inverse multiplexors.

These are only terminology and the miniture electronic components that
make up these units.

Best Regards,
Crystal.

	* Multiplexors:

	  Combines several narrow band width, into 1 larger bandwidth,
	  and vice-versa at the other end.
	  eg: 1000Mbps + 1000Mbps + ... > Multiplexor > 10000Mbps =>
	      10000Mbps pipeline =>
	      de-Multiplexor > 1000Mbps + 1000Mbps + ...

	  Multiplexors use industry standard bandwidth. So different 
	  brands can be paired over a pipelink. However, 
	  in practice, this is not true because the firmware or 
	  (more likely) the software drivers are not compatible.
	
	  There are other stuffs like cable connectors (coaxial/fibre
 	  optic) interfaces. 
	  And other things like onboard RAM cache, etc...

	  And this rule, although a maxim in times past, has been 
	  true recently, as more manufactures are using 
	  flash-upgradable firmware or software drivers, different
	  cable but with same speed performance, line extenders etc.

	  "Sub dividing" previous units, so to speak.
	  The reason is simple: by just chamging one aspect of a 
	  massive unit, impressive performance is achieved. 
	  This is known as: productivity/efficiency/
	  higher tech level :)

	* Concentrator:

	  Similiar to multiplexors but the "de-multiplexor unit" at
	  the other end, is part of a larger unit that is usu. 
	  embedded in the firmware (eg: a concentrator-to-switch).

	  If the de-concentrator function is a plug-in PCB card in a
	  switch, the manufacturers may simple call the two 
	  multiplexors.

	  In practice, concentrators belong to the old "SNA-IBM" era,
	  where multiplexor functions were built-in a controller unit 
	  that had a link to the mainframe. You attached a 
	  "concentrator" (actually a multiplexor) to a 
	  "controller" (that had multiplexor firmware as extra 
	  features).

	  Nowadays, concentrators are used in LANs (eg: Token Ring)
	  connected to a switch/server but you'll bet there is 
	  firmware "multiplexor function" in the switch/server 
		  - ending in a dedicated port on the casing; OR
		  - as a plug-in PCB card
		    (which may be called a "terminal adapter" :) ).  
	
	* Compressors: 

	  Compress data using firmware/software algorithms
	  eg: 1000Mbps > Compressor  > 800 Mbps => 
	      800 pipeline => 
	      de-Compressor > 1000Mbps

	  Data comprssors are like modems with Pkzip or MPEG-3 
	  compression schemes that gets rid of redundant data, 
	  eg. coding these into a smaller reference format, 
	  then sends these with the other non-redundant data 
	  to a compatible compressor.

	* Inverse Multiplexors

	  This is the reverse of multiplexors.
	  From one-big to many-small (and vice versa)
	  eg: 10000Mbps > Inverse Multiplexor > 
	      1000Mbps x 10 channels =>
	      10 x 1000Mbps pipeline =>
	      Inverse Multiplexor > 10000Mbps

	  Its simply a way for links that don't have a wide bandwidth
	  but only have narrow-bandwidth. This may be costs, or 
	  actual technology not available at the local premise.

	* Front-End Processor (FEP)

	  In the "old IBM mainframe" era, an FEP is simply a 
	  multiplexor dedicated for remote communications to 
	  branch offices. A pair if modems connects to the "FEP"
	  at the mainframe host, and a "controller" 
	  (actually a multiplexor) is at the remote site.

	  Bcos the mainframe has many dumb-terminal connections,
	  a "FEP" was usually a mini-computer that frees up CPU
	  processing in the mainframe, taking over the remote comms
	  function, and only passing "pure data" to the mainframe.

	  We must remember the mainframe topology is a centralized 
	  one. When PC computing power increased, due to more 
	  efficient semi-conductor technologies like 
		- VLSI (Very Large Scale Integration); 
		- ULSI (Ultra Large Scale Integration); 
		- low-micron; 
		- high clock speeds (Mhz/GHz) etc.;

	  it became no longer necessary for a centralized network.
	  
In fact, if silicon pathways are replaced with "optical linear
non/isometric chips" like described in the Star Trek manual, processing
speeds would sky-rocket. 

It would just be like connecting billions of Cray-3 into a PC. ;)
For example, one Cray-3 now has 2000+ Alpha CPUs. Multiply that by the
power of 10's at light speed...

This is, of coz, essential for stuffs like 
	* AI 
	  (the number of transitors circuits equals 
	  the number of biological neuron connections 
	  in a human brain);
	* 3-D
	  (3-D hologram in real space, eg: holodeck);