That's some list. Many of the questions are answered within the first couple screens of the forum. Are you just curious, or do you have a topology test coming up ?
Just to get things rolling....
Ethernet is a bus-based, connectionless protocol typically operating over a star topology (over UTP and fiber mostly these days). It operates asynchronously, transmissions can happen at any time.
Token Ring is is a token-passing protocol (so was Arcnet). Only the station holding the electronic token is allowed to transmit. Token ring was also usually implemented as a star topology. Token Ring is deterministic...meaning that if you knew all the components in the system, you could predict how much time it would take to get a token completely around the ring, hence predict traffic loads and host waiting times. Token Ring, as a defined standard protocol is much more robust and resilient networking protocol than basic standard Ethernet.
A "BUS" as it relates to Ethernet means that logically (if not electronically) all the stations are connected in parallel...as it relates to ATM networking, "BUS" is the "Broadcast and Unknown server," part of the LAN Emulation (LANE) standard used in ATM networking (ATM = Asynchronous Tranfer Method, not the cash station). The ATM BUS is the services component of LANE that emulates the broadcasting found in bus-oriented protocols. Since ATM is more like a phone system (each station must call and connect to the destination host) when a station tries to connect to another host that is unknown on the network, the BUS transmits the request via a one-to-many virtual circuit to all of the hosts it does know about. If a host tries to broadcast (like, an ARP for example), the BUS sends the request to all known stations. Many of the ATM vendors short-cut this step to preserve bandwidth by allowing the BUS to peek into the LES (LAN Emulation Server)tables which map the MAC to ATM address.
An Ethernet hub is essentially a group of repeaters in the same box. Every pulse is re-transmitted to every other port (including bad signals and packets).
An Ethernet switch is basically a multiport bridge. The most frequent implementation allows for several host-to-host conversations to occur at the same time, multiplying the apparent bandwidth. A switch does absolutely no good in a same-speed many-to-one access scenario..if fact, it probably slows things down a bit (generally speaking, some advanced features -not-typically found on consumer switches would cause be exceptions to the previous statement). Broadcasts still pass port-to-port (again, some switches will limit broadcasts...not typical on consumer-class switches).
Routers are used to connect network segments with different network addresses or protocols. Another benefit is that they (usually) DO NOT propagate broadcasts. They do alot of other stuff these days, but the basic functionality of a router is segmentation (limit the broadcast domain), and forwarding (or not) from segment to segment where the addresses or protocols are dissimilar.
A Broadcast Domain is the bountry of how far a broadcast packet can travel. In a hub or switch, the broadcast domain is essentailly the scope of the segment. If the entire switch or hub is defined as one segment, then that switch or hub is the broadcast domain. If the switch or hub is connected to another switch or hub directly, then the two switches or hubs are in the same broadcast domain. If the two switches or hubs are connected through a "standard" router, then there are two broadcast domains. If the router has an "IP Helper" function, then SOME broadcasts are forwarded, usually aimed at a specific host (as a Unicast).
This is also a reason the a "WINS" server may be needed for Microsoft networking. The WINS server keeps track of the Host's "NETBIOS NAME," and the associated IP address. NETBIOS (used in Microsoft LANMAN-style peer-to-peer networking) is not routable, so the packets are encapsulated (in TCP/IP or IPX/SPX packets). WINS allows the NETBIOS-speaking machines to find out who's-stuffing-what into IP packets, and indicates what IP address you have to transmit to if you want the packet to go to that particular NETBIOS name.
You can negate the need for a WINS server by listing the hostnames and addresses in a file called HOSTS or LMHOSTS (in the C:\WINDOWS directory for 9X, or C:\WINNT\SYSTEM32\DRIVERS\ETC directory for NT or 2K).
Star topology is much like it sounds, a central device with radiating lines terminating at the host or other networking device. It's considered (by some/many/most/all) to be superior to bus-oriented topologies (coax, like 10BASE-5 and 10BASE-2 Ethernet) because a failure at one "spoke" usually does not affect the other spokes. In the old coax days of Ethernet, if someone busted a connector or cable, the entire network (or significant portion) went down.
The MSAU (Multi-station Access Unit) is the device that makes the Token "Ring" into a star-topolology. The original units were a system of relays, activated by a "phantom" voltage expressed on the cable by each host. Later versions replaced the mechanical relays with electronic ones.
"Fast Ethernet," 100Megabit Ethernet uses Category 5 cabling. A group also came up with a spec to run 100M Ethernet over Category 3 cable...called 100BASE-T4 (I think)...and yet another group came up with "AnyLAN VG" which was similar. The basic concept was to split the signal and send one quarter of it down each of the four pairs...to be re-integrated at the receiving device.
Ethernet (any speed) is OK for basic data transfer. Token Ring is OK for basic data transfer, ArcNet is OK for basic data transfer (but watch those address assignments)...they're basically just "dumb pipes."
ATM is far superior, built from the ground up to handle data, video (real-time), and voice concurrently. It is the most scalable of the bunch, and doesn't require the likes of Spanning Tree for multiple parallel runs. In a couple years, when the entire MPLS spec is finally ratified and implemented, Ethernet will be where ATM was about five years ago.
Ethernet pors & cons....
PROs : It's cheap, it's easy to understand and use, it's everywhere, and the Internet presents itself to most users as some form of Ethernet (though much of it is carried in the cloud as ATM or straight SONET).
CONs: It's not always as easy as it looks, it's not easily scaled (parallel connections require additional management protocols, like Spanning Tree or MPLS), it doesn't lend itself well to multi-service use (like video and voice), there is no real "Class-of-Service" or "Quality of Service" (RSVP only guarantees a single span, not end-to-end, like ATM), there is no predictable delay variation time, and, it's so easy to use (or appears to be) that people don't bother to follow many of the "rules" or conventions for Ethernet.
Hope this helps. Next time, read the book though...or I'll tell your instructor you's slacking off and playing on the computer instead of studying.
FWIW
Scott
