Coordination strategies: Glossary

This is a brief glossary of terms; if there is something used in the main body of this white paper you don't understand, let me know and I'll add it here.

Directed multicast
A strategy in which a single host can issue an unreliable (in the sense of best-effort, UDP or IP-based) broadcast to only a small number of selected hosts. Described more thorough in RFC1458.
Latency
The time it takes to actually get data over a network once it is requested. This is typically up to a couple hundred milliseconds on fiber-optic-based WANs such as the Internet, although certain transcontinental hops are based on geosynchronous satellites, which increase latency up into the thousands-of-milliseconds range due to lightspeed delay. LANs usually have latencies in the single milliseconds or tens of milliseconds, and tend to exposure the underlying speed of the workstations involved as an important part of this latency. Systems which communicate using electronic mail instead of near-real-time TCP connections can have latencies ranging from minutes to days. Latency is important when considering coordination strategies amongst agents; if a great deal of interagent messaging is required to accomplish a task, and the user expects rapid response to queries, such messaging cannot take place either across particularly slow Internet connections or via email. Bandwidth and latency are independent quantities; it is possible to have a very-high-bandwidth connection that nonetheless displays high latency. An example would be a Cessna full of DAT tapes, which displays terabyte bandwidth but latencies measured in hours.
Bandwidth
The amount of data received per time unit (usually seconds) over a connection. Typical Internet bandwidths range from 14.4Kbps (a typical cheap modem of the times) to 45Mbps (a T3 link that's being completely utilized), which is a range of over three orders of magnitude. Certain experimental situations push this in both directions, to at least five orders of magnitude. Bandwidth and latency are independent quantities; it is possible to have a very-high-bandwidth connection that nonetheless displays high latency. An example would be a Cessna full of DAT tapes, which displays terabyte bandwidth but latencies measured in hours.

Previous | Table of contents


Lenny Foner
Last modified: Fri Dec 15 10:59:21 1995