It's not as simple as it sounds, of course. ;-) A good start might be Cisco's Technology Overview here: http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/index.htm That document shows typical header lengths for Frame Relay, ATM, ISDN, PPP, HSSI, SMDS, SDLC, etc. Search the group study list for more on Cisco's HDLC header. The HDLC frame format has been published here a few times. The exact header lengths will depend on which variety of each protocol you're using. For example, ATM headers are longer with AAL3/4 than with AAL5?? (I'm not an ATM expert.) Frame Relay fragmentation headers are different than vanilla Frame Relay headers. You probably would want to account for other protocol info that is overhead: call setup, signalling, SLARPs, LMIs, etc. Then, lest we forget, the equation depends on how much payload the upper layers put into frames. HTTP typically uses 400-600 byte packets, as do many database applications. Voice uses 64-byte packets. Terminal applications (like Telnet) use 64-byte packets. (A user types the letter A. The letter A is echoed back, and the echo is acknowledged.) Large frames (1000-1500 bytes) are typical of file transfer activities, such as saving files, executing applications, loading word processing documents, using FTP, etc.. E-mail is pretty impossible to predict. Attachments would use large frames. Message are hopefully small, but sometimes they are like this message. ;-) Then how about upper-layer overhead such as TCP acknowledgements? The number of ACKS depends on window size. Any retransmissions? And what else uses bandwidth? Routing protocols? Network management? Tunneling? RSRB? So.... Maybe just believe the SE who told you to assume about 25% is overhead, unless you want to do a real traffic analysis? ;-) 25% seems believable to me. How expensive is bandwidth? A real analysis might be worth the work... Sorry for all the hand-waving, but I'm in a bit of a rush. Hopefully you see the issues though. Priscilla At 04:36 PM 6/26/01, John Neiberger wrote: >Ah, this is one of those cases where I wasn't really answering the >question you were asking. :-) Sorry about that! > >I know I have some information like that around here but I can't find >it at the moment. Perhaps some of the listmembers that are more versed >in network analysis than I could provide some links. If I can find the >links that I've used in the past for this type of information, I'll let >you know. > >Regards, >John > > >>> [EMAIL PROTECTED] 6/26/01 2:04:07 PM >>> >You're comparing apples to oranges. A 10BaseT LAN is a shared medium, >which explains the oft-quoted-but-not-quite-accurate 40% max usage >figure. The theory is that with CSMA/CD, as traffic increases so do >collisions, which forces retransmits. This is exacerbated in a half >duplex environment. > >Dana's Comment - I realize that LAN and WAN environments are different. > I >was just using the 10Mb quote as an example. > >WAN links, such as a T-1, are synchronous (or isochronous or >pleisochronous or some other x-ochronous word that I don't understand) >full duplex connections. This means that frames are travelling boths >directions over the link at the same speed regardless of the amount of >traffic to be carried. In the case of a T-1, if you have 1.536 Mb/s >of >data to send, then go for it. The bandwidth is there if you need it. >For you hair-splitters, I'm purposefully not getting overly detailed. > >Of course, you have to factor in packet headers in your calculations. >Any data you have to send has to be encapsulated first. If you're >using >IP, then any data packets must be encapsulated with a UDP or TCP >header >and then with an IP header. This packet then must be placed inside >whichever datalink frame you're using, whether it's Ethernet, HDLC, >PPP, >or whatever. All of this creates overhead that you have to take into >account when calculating how much bandwidth you actually have >available. > >Dana's Comment - On WAN links, I just was wondering how what percentage >I >should give to headers, framing, signaling, etc. One Cisco SE told me >that >I should calculate 25% to cover those issues. I.E. On a 512K link I >could >only expect 384K of usable bandwidth. I figured that different >technologies should have different percentages - ATM vs. ISDN vs. Frame > >Relay etc. That is what I am looking for is specifics per technology >as >opposed to a general WAN figure. I hope these notes clarify my >question. > >But the moral of the story is that point-to-point WAN links suffer >from >different issues than CSMA/CD networks. > >Dana's comment - John, thanks for your comments. > >Regards, >Dana >CCNP, CCDP ________________________ Priscilla Oppenheimer http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=10017&t=10017 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
