Ismail Butun Effects And Causes Of Latency Spring 2008
Effects and Causes of “Latency” Submitted by ISMAIL BUTUN Department of Electrical Engineering University of South Florida An abstract report submitted in partial fulfillment of the requirements for the course EEL 6935 – Wireless Network Architectures and Protocols (Instructor: Prof. Ravi Sankar) Note: The surveys are achieved on 1997, and they don’t demonstrate today’s theoretical and practical achievable limits. Spring 2008
In communication systems latency is an important issue to think about. For example, the round‐trip delay over a telephone network should be less than 100ms, which is a well known rule in the telephone industry. If the latency is more than 100ms, then human ear can easily detect the gap in the conversation. As it is observed in communication over satellites, in which the latency is about 250ms, the conversation becomes faded and interrupted because the 100ms limit is exceeded. The latency consists of several reasons in the overall communication system • MODEM latency • Network (queuing in the routers) latency • Communication channel latency One of the main reasons of the latency in communication MODEMs is perception. Hence the main focus of the whole communication industry is focused on large bandwidth and high throughput, perception is ignored and comes up with drastically results on the overall communication latency. With traditional modems the round‐trip delay time is around 260ms. This is caused by data compression, automatic error correction (forward error correction) and buffer time (waits until 100bytes of data are collected). Those modems have old design criterions for transmitting documents only and are not designed for interactional communication. NETWORK latency, is also called queuing delay which is caused by the congestion of the network routers. It is considerable to aim the COMMUNICATION CHANNEL latency as 70ms for interactive communication, which is not easily achievable with current communication networks. The least amount of latency can be achieved with glass fiber, and is not less than 200ms for long distance communication (for a round trip and the distance is assumed to me 40000km). It is the latency just only through the communication channel. Large bandwidth doesn’t mean low latency at all. There is no guarantee of low latency comes with large bandwidth guarantee. So latency also has to be considered through the implementation of the communication system. There is nothing to do to decrease the channel latency, owing to the fact that we are sending out data at the practical achievable limits, namely the speed of the light. To decrease the network latency, sending packets according to their priority should be a solution. There is much more time spent through the MODEMs compared to communication channel. Thus to cope with latency, the latency hiding techniques are inevitable to use which provide us illusion of interactivity. Dead‐reckoning which is smooth exploration of data can also be used to cope with latency virtually. References  S. Chesire, Latency and the Quest for Interactivity, http://www.stuartcheshire.org/papers/LatencyQuest.html, Nov. 1996.  S. Chesire, Its the Latency, Stupid, http://www.stuartcheshire.org/rants/Latency.html, May 1996.  S. Chesire, Latency Survey Results, http://www.stuartcheshire.org/rants/LatencyResults.html, 1997.