On Fri, 19 Apr 2002 23:54:49 -0500 Timm Murray <hardburn at runbox.com>
writes:

> If it seems like they are always receiving data, that is because 
> they are.  
> Below is a quick introduction to how networking works.
> 
> There are two layers we are concerned with here: (1) Physical and 
> (2) Data 
> Link (there are five others in the most common networking model 
> (OSI), but 
> these are the important ones for our discussion).
> 
> The Physical layer is basicaly a wire (or fiber or radio waves or 
> whatever) 
> and the bits that travel over it.  Nothing particularly 
> inteligent--it just 
> provides a way to make electrons (or photons) from one place to 
> another.  
> Repeaters and hubs are layer 1 devices, as well.  This part is done 
> purely in 
> hardware (usually).
> 
> The Data Link layer creates some logical orginization to a network.  
> You have 
> some sort of address (for Ethernet, this is a MAC address), but no 

Hehe, if not for my class at school I would have thought you were talking
about Apple Macintosh and my head would be spinning in confusion. :) 
But, I know what you mean. :)  Our system is setup to take the MAChine
addresses (ie 17204a) and give them IP addresses and such through DHCP. 
Very nice of it to do that. :)  Yesterday I had to put my removable lab
hard drive in a different computer than I was useing when I installed
win98, winnt4, and win2kpro so it bombed, sort of, because a computer
with that exact address was already on the network.  So I said "forget
this" and I named my machine "Krepta" on all three OSes, cause I KNOW no
other machine will EVER have that address. :)

> actual 
> routing takes place.  The addresses are just there to provide some 
> way to 
> make each computer unique on the network.  NICs, bridges, and 
> switches are 
> layer 2 devices.  This part is partialy done in software, but is 
> done mostly 
> in hardware.

Ah, ok. :)

> 
> So why are your computer's constantly receiving data?  Well, think 
> of the 
> situation where you need to run Ethernet beyond it's 100 meter 
> limit.  You 
> get a repeater (a layer 1 device), which has the sole job of taking 
> the 
> signal in one end, strengthening it, and spewing it out the other 
> end.  The 
> two end points on the system might as well be on the same wire.

Makes sense, all broadcast systems use signal repeaters to keep the
signals from degrading too much.

> 
> To get a hub, you basically stick a bunch of repeaters together; in 
> other 
> words, hubs are just multiport repeaters.  So, all computers 
> connected to a 
> hub might as well be connected to a single wire.

Yeah.  That makes sense too. :)

> 
> Now you ask "since they are all on the same 'wire', what happens 
> when two 
> machines try to talk at once?"  Answer:  You get a collision.  In a 
> collision, the signal is destroyed bit-by-bit.  All devices on that 
> "wire"  
> notice this and stop sending data for a random ammount of time and 
> then start 
> resending.  In general, the more devices you have on that hub (or 
> "collision 
> domain"), the more likely collisions will happen, thus reducing your 
> total 
> throughput (because the collisions stop everyone from sending 
> stuff).

So thats why the network sometimes just goes down, or parts of it anyway.
 Signal collisions!!!  Jeeze. :(

> 
> (To the gurus:  Yes, I know I'm ignoring Token Ring networks.  
> Nobody cares 
> about Token Ring anymore, so just shut up :)

Is Token Ring better?  If so, why aren't people useing it?

> 
> Taking our example further, let's say you have two hubs that you 
> want to 
> connect together.  You are worried, however, that this will make 
> your 
> collision domain much too large.  No problem, just use a bridge.  A 
> bridge, 
> being a layer two device, knows something about the addresses being 
> used on 
> each side.  If machine A sends something to machine B, and they are 
> both on 
> the same hub, the bridge knows that it shouldn't forward the traffic 
> onto the 
> second hub.  Thus, the bridge has split your collision domain, 
> otherwise 
> known as "segmenting".

Ah yes, switches, thats what I was trying to think about to reduce
traffic over all the lines.

> 
> Remeber how hubs are really just multiport repeaters?  What happens 
> when you 
> create a multiport bridge?  You get a switch.  Switches basicaly 
> saved 
> Ethernet, because without them, collision domains can quickly get 
> too large 
> to handle.

Yes, I see, very very good idea to use bridges(switches).

> 
> Like a bridge, a switch will remember what MAC addresses are on each 
> port, and 
> are thus able to make forwarding decisions.  Switches are usually 
> more 
> expensive than hubs, and create slightly higher latenices (since 

What is a latinice?  Do you mean Latency?  I've noticed that word being
used on Network Games like X-wing Vs. Tie Fighter, but I didn't know what
it meant, other than signal speed or something.

> they 
> actually have to do some work), but throughput is increased to 
> near-maximum 
> for each switch port.

How about hooking in some kind of RAM that can be used as a buffer inside
a switch so that it will forward the data when a machine is ready to
recieve it?

> 
> The usual network configuration is to use hubs on the workstations, 
> then 
> connect those hubs into switches (servers are often, but not always, 
> directly 
> connected to switches).  If the orginization can afford it, it will 
> get the 
> Nirvana of Networking . . . that is, having a network which has no 
> hubs at 
> all.  This is called a "microsegmented" network.

Yahoooo!!!!  I would love to hook up a massive network of nothing but
switches, no hubs!!!  But it still seems to me like there would be WAY
too many "Shared Resources" on a serverless network, even a
"microsegmented" one, to be any real use to any large organization.  I
would still really like to put Freenet nodes into every PC to make the
entire network act like one massive server.

Of course, a serverless network might still need to be created in a kind
of higharchy or something.  Hmmm. :(

Would DHCP automatic IP addressing be able to work on a "microsegmented"
network just like on a HUB network?

> 
> There is probably a lot more information up there then you really 
> wanted to 
> know, but I hope I answered your question somewhere in there :)

Oh don't say that, I want to know as much as possible, thats why I'm in
the "Computer Networking and Information Technology" program at High Tech
Institute. :)

________________________________________________________________
GET INTERNET ACCESS FROM JUNO!
Juno offers FREE or PREMIUM Internet access for less!
Join Juno today!  For your FREE software, visit:
http://dl.www.juno.com/get/web/.

_______________________________________________
Chat mailing list
Chat at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/chat

Reply via email to