Let me expand on Priscilla's thoughts. The reality is that for more than trivial problems, you need a lot of information to model latency. Bandwidth is easy.
At 1:24 AM +0000 11/23/02, Priscilla Oppenheimer wrote: >Mirza, Timur wrote: >> >> does anyone have a good reference (e.g., white paper) on the >> nature of >> bandwidth vs latency & the distinction bet/ the two? > >Well, the distinction is easy. They don't mean the same thing at all. A good >site is Merriam Webster's online dictionary. The 2nd definition for >bandwidth is: > >Bandwidth: the capacity for data transfer of an electronic communications >system > >Latency, on the other hand, means delay. Websters isn't too helpful in this >case, but might help you understand the origin of the word, which is related >to dormancy. > >Cisco's Terms and Acronyms document has a couple definitions of latency, >which are somewhat helpful: > >1. Delay between the time a device requests access to a network and the time >it is granted permission to transmit. > >2. Delay between the time a device receives a frame and the time that frame >is forwarded out the destination port. > >Those definitions allude to the many contributors to delay (latency) on a >network: > >* media access time >* queuing time at internetworking devices >* processing time at internetworking devices and at the sender and receiver >* serialization delay to send and receive bits at the rate specified by the >bandwidth of the sending and receiving interfaces >* propagation delay which is distance dependent and to a certain extent >medium dependent, although most media support about 2/3 the speed of light > >Testing latency is reasonably easy. Just do some pings. Predicting, >modeling, and simulating delay is advanced engineering. A few books cover it >at a very basic level, including Top-Down Network Design by Oppenheimer, and >Data Network Design by Spohn. > >Howard Berkowitz has written some RFCs that discuss performance measurement, >if I recall. My specific work (almost at RFC -- the IESG has told us what editorial changes are needed for RFC approval) deal with convergence, which takes place before data transfer can happen. http://www.ietf.org/internet-drafts/draft-ietf-bmwg-conterm-03.txt I have done delay work with ANSI/CCITT/ITU, in ANSI X3.102 and X3.141 (probably not online), and in CCITT X.140. There are two relevant IETF working groups, BMWG and IPPM. BMWG deals more with device delay and IPPM more with delay between devices. BMWG approved RFCs, of which 2544 is a good starting place. It's also worth looking at the I-D's. Benchmarking Terminology for Network Interconnection Devices (RFC 1242) (22349 bytes) Benchmarking Methodology for Network Interconnect Devices (RFC 1944) (66061 bytes) Benchmarking Terminology for LAN Switching Devices (RFC 2285) (43130 bytes) Terminology for IP Multicast Benchmarking (RFC 2432) (29758 bytes) Benchmarking Methodology for Network Interconnect Devices (RFC 2544) (66688 bytes) Benchmarking Terminology for Firewall Performance (RFC 2647) (45374 bytes) Terminology for ATM Benchmarking (RFC 2761) (61219 bytes) Benchmarking Methodology for LAN Switching Devices (RFC 2889) (73251 bytes) Methodology for ATM Benchmarking (RFC 3116) (294857 bytes) Terminology for Frame Relay Benchmarking (RFC 3133) (44182 bytes) From IPPM, Framework for IP Performance Metrics (RFC 2330) (94387 bytes) IPPM Metrics for Measuring Connectivity (RFC 2678) (18087 bytes) A One-way Delay Metric for IPPM (RFC 2679) (43542 bytes) A One-way Packet Loss Metric for IPPM (RFC 2680) (32266 bytes) A Round-trip Delay Metric for IPPM (RFC 2681) (44357 bytes) A Framework for Defining Empirical Bulk Transfer Capacity Metrics (RFC 3148) (36041 bytes) One-way Loss Pattern Sample Metrics (RFC 3357) (30570 bytes) IP Packet Delay Variation Metric for IPPM (RFC 3393) (47731 bytes) I also have practical discussions in my books from Wiley, _Building Service Provider Networks_ and _WAN Survival Guide_. Also see _ISP Survival Guide_ by Geoff Huston. > >There are graduate level computer science classes that cover performance >measurement in computer networks at many universities. > ANd, unfortunately, when you start dealing with it at more than the link level, you get into this level of sophistication. Various sources for Internet-wide behavior include the work of Laibovits and Ahuja, which you can find with search engines or by searching through NANOG.org. There's good work at CAIDA.org. A search engine should give you Vern Paxson's dissertation "End-to-End Packet Dynamics in the Internet," and also take you to Sally Floyd's homepage. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=57946&t=57899 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
