SMAC May 1999 Durkin - 6

CHART A

Percentage of Usable Bandwidth – CSMA/CD vs. SMAC

1 0 Mbps 100 Mbps 1000 Mbps SMAC at 10, 100, CSMA/CD CSMA/CD CSMA/CD and 1000 Mbps

4096

bits per frame with 200 – 300 nodes

Although this is an impressive market share, CSMA/CD – no matter how much it's been tweaked and improved over the years – still has its one fatal flaw: collisions. As shown in Chart A,CSMA/CD is not cost-effective or practical on gigabit speed networks for the following reasons: Data latency – Because CSMA/CD is, quite literally, a hit-or-miss protocol, it is impossible to guarantee when and how much data will be transmitted. This makes it practically impossible to distribute real-time priority data packets. Real- time priority processes include financial transactions and broadcast-quality videoconferencing. Quality of services – The potential speed of a Gigabit Ethernet becomes irrelevant if it can't deliver consistent throughput – which it can't. The CSMA/CD paradigm puts Ethernets at an extreme disadvantage, at any speed, in competing with technologies such as ATM, SONET, and Fibre Channel. Circuit-switching – With circuit-switching, network managers can build large networks in which various data packets may be intelligently switched between Ethernet segments. CSMA/CD places a difficult burden on the network manager because of its random, unsynchronized nature. The difficulty lies in how to determine what circuit-switching assignments would be optimal. Packet size – According to the Gigabit Ethernet Alliance, Gigabit Ethernet networks require data frames to be of static size, regardless of the amount of information in the frame. The SMAC protocol can transmit frames of variable
   