Interesting problem. I think that I would look at QoS options or an application layer solution. Perhaps CAR configured for a maximum rate that would force the application to not have sufficient B/W on the path?
--- Darrell Newcomb wrote: > I try not to use the below logic on my networks, but > have also never had > it fail to deliver service when there was no other > choice. > > The common streaming of windows media and real have > such large client > side buffers that you'll find you can seemingly > overload the link > without having any user observable qualitative > difference. Some factors > which contribute even more to the success of > overloading are the bit > rate varies as the encoders don't always output the > maximum data rate. > The fact that most streams on the public internet > are short lived, the > standard buffers can cover the end of the stream the > user is still > viewing leaving capacity for other streams to go > through their peak > startup period. The traditional stat muxing factors > come into play > where depending upon the application there is some > downcycle in > streaming usage in the workflow. You only need a > 2.5:1 to get 300kbps > streams through uncongested. > > Lastly I think you are approaching the wrong > problem. Non streaming > uses for the same 2Mbps link will be the big enemy > of predictably good > streaming performance. Your application may even be > one of those by > downloading other supporting data... > > To more directly approach the problem space you > posed: > -There is xauth in pixOS and I believe IOS NAT > -Couple that with a creative authentication server, > or script to control > it.... > -The above should get you the max number of sessions > through. > -Can't recall the reflexive access lists with CAR > ball of wax off the > top of my head. But there is some per-session rate > limiting in cisco. > > There are various rate limiting equipment out there. > Riverstone has > good affordable routers for this, Netscreen claims > to do it(haven't used > them yet), and Packeteer also does this type of > thing. There is more > but I believe them to be the notables. > > There are proxy and/or cache products which would > address the max number > of sessions issue and maybe address the usage > pattern you have. > > Not that I'd recommend this, but if your application > and rest of the > network path can adequately support forcing the > streams over a tcp > session you'll probably find it much easier to deal > with the rate > limiting. But really try to handle it without > forcing tcp as any > backoffs will hurt the qualitative performance if > there are other > signficant numbers of tcps over any congested > link.(read: IME(nee > opinion) tcp will backoff quicker than a given > streaming protocol) > > Good Luck, > Darrell (always looking for contract work) Newcomb > > Gaz wrote: > > > > Hi all, > > > > I'm after some ideas if you'd be so kind :-) > > > > A 2Mb link being used mainly for streaming media > has about 15 potential > > users. The task is to limit the number of users at > any one time to four, so > > they have half a Mb each (ish). > > > > My initial idea, which I must admit, I dont think > is such a good one is to > > set up a NAT pool of four addresses, and drag the > translation timeout down > > to about a minute (yet to be tested), so that the > first four users to pass > > traffic will be translated and allowed through, > but after that, they'll > have > > to wait. > > > > I'm off to look at something like TACACS to see if > I can control network > > authorization by number of users (shot in the > dark). > > > > No equipment in place yet, so we have a clean > drawing board. > > > > Anybody have any neat ideas please!!!!!! > > > > Thanks, > > > > Gaz [EMAIL PROTECTED] ===== Robert Padjen __________________________________________________ Do You Yahoo!? Great stuff seeking new owners in Yahoo! Auctions! http://auctions.yahoo.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=33818&t=33306 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
