Class a jitter scenario Using the Simula rpr 3 Java simulator

Class A jitter scenario

• Using the Simula RPR V2.3 Java simulator

• Students implementing the scenario:

• Bjornar Libaek
• Petter Teigen

• Based on “cls09_fairness1.pdf” dated june 23. 2003 from Dave James

• In RPR version 2.3, the D shaper does not shape the transit traffic. I believe this results in the class A jitter shown. We are in the process of changing the simulator so that shaper D will shape the STQ traffic. Results from these simulations are hopefully available tomorrow (June 31. 2003)

Scenario (From Harry Peng)

• 2000 km between S30 and S31

• 1 km between S1 and S2, S2 and S3, …. S29 and S30

• 1 km between S31and S32, …., S61 and S62

• 5 us per km

• 1000M lineRate

• Total distance: 1940 + 29+ 31 = 2000

• STQ size = 1% off Ring circumfernce

• = .01(5e-6 [s/km] * 2000 [km] * 1e9 [bit/s])
• = 100 kbits or 12.5 kilobytes

• S61: b) Dual Queue design

Additional scenario info.

• S0 to S30 are producing at full speed

• S31 to S60 clients are producing A1 traffic at 3% of line rate.

• S61 client is producing A0 traffic at 1% of line rate

• All Mac clients have small add buffers (two frames).

• Two STQ sizes: 12.5 Kbytes and 25 K bytes

• Other RPR parameters are defaults.

The next 6 pages show

• Class A1 traffic from S 31 to S60 (STQ = 12.5 Kbytes)

• Class C traffic from S1 to S30 (STQ = 12.5 Kbytes)

• Class A0 traffic from S61 (STQ = 12.5 Kbytes)

• Class A1 traffic from S 31 to S60 (STQ = 25 Kbytes)

• Class C traffic from S1 to S30 (STQ = 25 Kbytes)

• Class A0 traffic from S61 (STQ = 25Kbytes)

Conclusion

• All class A1 traffic has a long period from time 12 ms in which it is not allowed to send.

• With 12.5 Kbyte STQ size this lasts about 20 ms, with 25 Kbyte STQ size this is lasts about 50 ms.

• Class A0 traffic from S61 to S62 behaves as the class A1 traffic.

• Shaper D running negative is probably the cause for the class A1 delay.

• Stay tuned for results when shaper D shapes STQ traffic.