From dumb repeater to Superhub

The dumb repeater is an integral part of Ethernet. The
little grey box has evolved over 20 years, still rules the networks.

How it works


The repeater helps expand the network by connecting multiple segments together. It is
"dumb" because it cannot process any data. It merely reduces signal noise and
retransmits it. Hence, it resides at Level 1 of the Open Systems Interconnect
(OSI)
network model.

The repeater receives the signal from a node. The quality
of the signal that is received depends on factors such as distance and medium. This signal
is digitally perfected, regenerated, and retransmitted to all the other ports. In the
process of propagates network traffic and collisions across ports and segments, it
introduces a small delay that affects much of the network design and diameter rules.

The avatars of repeaters

The repeater has taken several avatars over time. With the increasing popularity of
Unshielded Twisted Pair (UTP) cables, the repeater took the form of a hub. Thereafter, it
always remained the focal point for all the traffic in the network. As networks grew in
size, the repeater became a stackable hub. These are separate boxes, interconnected to
each other with a special stacking cable to form a "single hub". Technological
advances ensured that the overall stack delay was no more than the standard repeater
delay. This made it very convenient to start on a small scale and scale up later.

The advent of Simple Network Management Protocol (SNMP)
meant that the repeater was not "dumb" anymore. In addition to repeating the
signal, the additional intelligence helped provide information for network management.
Even then, the popularity of inexpensive "dumb hubs" exceeds the number of
SNMP-based "manageable" hubs.

When the Fast Ethernet standard was framed, backward
compatibility with 10 Mbps Ethernet was a major objective. Even important factors such as
network diameter were compromised. Fast Ethernet using repeaters cannot exceed a network
diameter of 205 meters. In fact, this restricted the use of 100 Mbps networks to small
workgroups alone. That explains the many 100 Mbps-capable PCs still running at 10 Mbps.

Dual-speed hubs

Although users with deeper pockets used switches to upgrade from 10BaseT to 100BaseT, most
others waited for a more acceptable solution. The dual-speed hub was one. Even though many
predicted that the advent of switches would kill hubs and repeaters, the dual-speed hub
has bridged a huge gap between 10BaseT and 100BaseT network systems. It heralded the
convergence of repeater and switching technologies.

Dual-speed hubs provided a bandwidth of 110 Mbps, and
effectively, two collision domains, further improving network performance. It resulted in
cost-effective migration to a full-100 Mbps network. For stacking a dual-speed hub, only a
single two-port internal switch is required in the entire stack.

At about the same time as dual-speed hubs came desktop
switches. These did not do very well. The dual-speed hub was the clear winner because on
account of its easy migration factor. The performance issue, along with migration, needs
more attention.

Hubs had a dramatic cost advantage over switches. The
challenge was to come up with switch-like performance at hub-like costs.

Most switches provide full-duplex bandwidth i.e. a 100 Mbps
switch would provide 100 Mbps in both directions simultaneously. This capability is fully
used only by high-end servers or ultra-high-performance workstations. In most other cases,
it is not utilized to its maximum capacity.

Enter the Superhub

Superhub, a term coined by Accton, is also called a port-switching hub by 3Com.
Essentially, it is a half-duplex switch. A 100 Mbps Superhub provides a total bandwidth of
100 Mbps i.e. full 100 Mbps in a single direction, or say, 70 Mbps in one direction and 30
Mbps in the other simultaneously. Like a switch, it allows independent switched paths.
Each node will get the bandwidth guaranteed to it.

Superhub is IEEE 802.3 compliant, and thus interoperable
with Ethernet and Fast Ethernet. In addition to the two, it offers simultaneous bridging
for all ports. Cheaper "bridging" techniques between each port replace costly
switching fabrics. Thus, simultaneous data transfers are possible. Each port has
auto-negotiation capability ensuring that 10 Mbps nodes and 10/100 nodes can work
together. In this case, the Superhub behaves like a switch, but it is a lot cheaper.

The Superhub advantage

Conventional hubs allow only one port to operate at the full 10 or 100 Mbps at any given
time. All users have to share the total available bandwidth. The Superhub gave what many
users desired: the convenience of 10 Mbps connectivity, the speed of 100 Mbps technology,
and switched performance. Furthermore, each user could have dedicated bandwidth allowing
all ports to simultaneously operate at 10 or 100 Mbps, and thereby eliminating congestion.

The limitation of dual-speed hubs were that a single
two-port 10/100 bridge had to be used to convert 10 Mbps traffic on one port to 100 Mbps
traffic on another. Superhubs work around that by simultaneously bridging dissimilar
traffic at full Ethernet speeds on each port. This feature even helps do away with the
205-meter hub diameter barrier.

With Superhub, users can migrate to cost-effective and true
100Mbps networks. The dumb repeater has come a long way. It fulfils the basic functions
required of it, and goes on to do much more.