Advise or help on 100baseT

I'll give you something to think over, did they upgrade anything else?

There's a reason why the industry no longer uses hubs. Hubs are simply signal replicators. Any signal on an input port is replicated to the output on all other ports. This means that in a network consisting purely of hubs, every device sees the traffic originating from every other device on the network regardless of how it is connected or how many hubs are between them. Switches on the other hand switch traffic from a source port to a destination port, traffic that is not destined for a particular port does not appear on that port and is not visible to other devices.

It should be apparent from the above why switches are preferable to hubs from a traffic management perspective. Traffic on a hub based network places the bus arbitration and collision detection requirements on the NIC whereas traffic on a switch based network centralizes these to the switch. Trust me when I say that a switch will do a far better job of managing the topology than a bunch of disassociated NICs.

In other words, in a hub based network each NIC sees all traffic from other NICs and must attempt to pick a timeslot in which the link is expected to be empty before it can transmit its own data. If two separate devices try and transmit at the same time, a collision occurs and they both back off for a random duration before trying to transmit again. As the number of devices attempting to transmit at the same time increases the probability of a collision on the initial and all subsequent retransmission attempts approaches 1 (guaranteed) and throughput falls to zero.

Switches use internal buffers to manage data in the best possible fashion. Since data is switched from one port to another rather than broadcast to all ports, collisions can only ever occur between two ports communicating with each other and this is easily designed around.

Furthermore, an NIC attached to a network through a switch negotiates a link speed with the switch itself. This means that a switch can handle a number of 10BASE-T, 100BASE-TX, and 1000BASE-T links at once on various ports any traffic between any two ports will be handled at the slowest speed between those two ports only. An NIC attached to a network through a hub negotiates a link speed with the entire network. If even one NIC on that mess of hubs is only capable of 10BASE-T despite 100BASE-TX hubs and NICs everywhere else, the entire network will fall back to 10BASE-T.

So what does this mean in practice? There's a practical limit the number of hubs and repeaters that can be installed on a network. I'll let you do the research about this though but given the above information you should easily be able to head in the right direction.

Good luck!

The drivers have nothing to do with this, I can assure you of that. This is a purely logical problem.

The network topology is also irrelevant. A bus topology may become the limiting factor in a switch based network when outbound traffic from multiple ports has to be time multiplexed by the switch but in a hub based network all traffic originating from a transmitter is broadcast to the receiver on every other NIC. Since hubs do not perform time multiplexing it does not matter if they're connected in a ring, a star, or a bus. When two devices attached to a hub (not a switch) decide to transmit at the same time, the signals interfere with each other and all receivers receive only an unintelligible mess of data (a collision). When a collision is detected, all transmitters involved in the collision will have to repeat the data that collided. They pick a random backoff period (using a PRNG) and try again some time later in the hope that the other transmitter(s) will not pick the same backoff period. At that later time they will not only be competing with other transmitters trying to send new data, but also with the transmitters with which they collided. If the number of transmitters trying to transmit data is sufficiently high, every single transmission will result in a collision and network throughput drops to zero.

What I deliberately did not mention above is that the recommended maximum number of Ethernet transceivers and repeaters that can be connected to a 100BASE-TX hub network is less than the recommended maximum number of Ethernet transceivers and repeaters that can be connected to a 10BASE-T hub network. So, simply replacing the hardware with ones that meet a new standard won't help unless the topology also meets that standard.

I addressed several reasons why they would not see a performance increase when moving from 10BASE-T to 100BASE-TX.

1. If one of the NICs is not capable of 100BASE-TX the entire network will operate on 10BASE-T even if all of the hubs and other NICs are capable of negotiating 100BASE-TX

2. If there are a large number of transceivers trying to talk at once on a hub based network (packet switched traffic is inherently bursty) they will jam up. A 24 port gigabit switch typically has 48 gigabits of switching power (24 ports * 1 gbps * 2 directions = 48gbps) but a 100 megabit hub can handle only 100 megabits per second of broadcast power from any single input to all outputs

3. There's a limit to the number of repeaters that can be installed between any two hosts on a hub network. There is no such limitation for switches. 10BASE-T allows for 4 repeaters between hosts, but 100BASE-TX allows for only two and only under very strict circumstances. If my assessment of your problem's topology is correct, each case involves no less than 3 hubs: one for the workstation group, one for the backbone bus itself, and one for the other workstation group or the servers. This setup would work for 10BASE-T but not 100BASE-TX. Setting up the network described using 100BASE-TX hubs would most likely fail.

I'm not sure what your teacher is looking for, but the above reasons are why it would not work in reality.