You should work with ratios rather than absolute numbers. At 06:19 PM 10/18/01, Jon Tucker wrote: >I'm looking for feedback that I am/not calculating my custom queue byte >count the correct size. > >Below I've made up a circuit with a 56K cir and a port speed of 64K. Assume >the carrier lets me burst to port speed. > >My first question would be do I calculate the queue sizes based on cir or >cir+be? I did the calculations below based on cir.
It doesn't matter. >I said I wanted to give 10% to WWW, 25% to DLSW, 25% to VoIP, and the >remaining 40% to everything else. > >56000 bits per second = 7000 bytes per second (56000/8) > >7000 * .10 = 700 >7000 * .25 = 1750 >7000 * .40 = 2800 This would work. But you could accomplish the same 10%, 25%, 25%, and 40% with: 1000 bytes 2500 bytes 2500 bytes 4000 bytes or any other combination with the same ratios. Think of the output interface as a conveyer belt carrying, oh let's say boxes of cereal. On each cycle, for every 10 boxes of Cheerios you place on the belt, you place 25 boxes of Wheaties, 25 boxes of Rice Crispies, and 40 boxes of Bran Flakes. Then you drink your morning espresso and for every 100 boxes of Cheerios you place on the belt, you place 250 boxes of Wheaties, 250 boxes of Rice Crispies, and 400 boxes of Bran Flakes. Each cereal still uses the same percentage of conveyer bandwidth compared to the other cereals, despite your hyperactivity. That's why the Ethernet example below works. They didn't worry about actual Ethernet bandwidth. To really do a good job with custom queuing, however, you do also care about the actual byte size that you use. It it's an FTP application, for example, it would be silly to use 25 bytes or even 250 bytes since each packet is probably 1500 bytes and packest don't get divided up. But 25 would be OK if it's VoIP. I hope that didn't just confuse matters. I bet it gave you something to chew on though. (bran flakes!) Priscilla >interface serial0.2 > frame-relay class pointtwo > >map-class frame-relay pointtwo > frame-relay traffic-rate 56000 64000 > frame-relay adaptive-shaping becn > frame-relay custom-queue-list 2 > >queue-list 2 protocol ip 1 tcp www >queue-list 2 protocol dlsw 2 >queue-list 2 protocol ip 3 list 180 >queue-list 2 default 4 >queue-list 2 queue 1 byte-count 700 >queue-list 2 queue 2 byte-count 1750 >queue-list 2 queue 3 byte-count 1750 >queue-list 2 queue 4 byte-count 2800 > >access-list 180 permit udp any any range 16384 16484 >access-list 180 permit tcp any any eq 1720 > > > >Going by a calculation I've read in the Slaterlee/Hutnik book (pg 481). To >assign 75% of all bandwidth to one queue and 25% of the bandwidth to the >other queue, they made two queues the first 7500 bytes and the second 2500 >bytes and placed that queue on an ethernet interface. Those byte count >numbers don't follow the bandwidth capacity of an ethernet interface. So, >how did those figures get calculated? > > >Another way I've seen these custom queues calculated: (in a training class >I attended last month) > >Using the figures from my test scenario of 10% WWW, 25% DLSw and VoIP and >40% default. Using the default byte size of a queue, 1500 bytes. > >My 10% queue would equal 1500 bytes. > >10%=1500bytes - WWW >25%=3750bytes - DLSw >25%=3750bytes - VoIP >40%=6000bytes - Default > >The first method I listed seems to make the most sense to me. > >This whole process has me confused as to which method is the correct method. >Any help shedding light on this would be extremely welcome. > >Thanks, >- JT ________________________ Priscilla Oppenheimer http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=23483&t=23449 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
