1 / 14

v1.0 - 20050426

Telecommunications Industry Association TR-30.3/08-07-012 Arlington, VA July 14 - 15, 2008. v1.0 - 20050426. TIA-921 improvements Chip Webb, CTO – Anue Systems July 2008 - Arlington. www.anuesystems.com. Outline. Areas for future work Bandwidth Limitation Duplicate test cases

tasya
Download Presentation

v1.0 - 20050426

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Telecommunications Industry Association TR-30.3/08-07-012 Arlington, VA July 14 - 15, 2008 Anue Systems, Inc. www.anuesystems.com v1.0 - 20050426

  2. TIA-921 improvements Chip Webb, CTO – Anue Systems July 2008 - Arlington www.anuesystems.com

  3. Outline • Areas for future work • Bandwidth Limitation • Duplicate test cases • Behavior with CBR versus VBR sources • Time correlation of forward/reverse model • Conclusions Anue Systems, Inc. www.anuesystems.com

  4. Bandwidth limitations • TIA-921 does not require policing or shaping of user data • This is wrong for several reasons: • In real-life, a rate-limited link drops packets when the user bit rate exceeds the link bit rate. • In real life, a rate-limited link drops packet bursts that exceed the available buffer size. • In real life, a rate-limited link smooths out traffic bursts. The amount of that effect will depend on link utilization and the proportion of user traffic that that represents. Anue Systems, Inc. www.anuesystems.com

  5. Example: Bandwidth smoothing effect Bursty Input Pkts p1 p2 p3 p4 p5 p6 p7 p8 time Egress Queue Input Pkts Output Pkts Note: P5 is dropped. Note: The output bit rate for this example is ½ the input bit rate. The effect is even more pronounced for larger ratios (such as T1/GigE). Smoothed Output Pkts p1 p2 p3 p4 p6 p7 p8 time Anue Systems, Inc. www.anuesystems.com

  6. Duplicate Test cases • The TIA-921 model is unidirectional • The code in the reference model ignores reverse link bit rates, except for calculating LOOs • Examples: • TIA-921: TC10 and TC19 are same (A2B) • TIA-921-A: TC49 and TC73 are same (A2B) • Summary: • TIA-921: 92 of the 120 have duplicates • TIA-921-A: 48 of 96 Lan2Lan cases have dupes. Anue Systems, Inc. www.anuesystems.com

  7. TIA-921 Duplicates (A2B) direction Anue Systems, Inc. www.anuesystems.com

  8. TIA-921-A Lan2Lan Duplicates (A2B) Anue Systems, Inc. www.anuesystems.com

  9. Reverse model correspondence • For the reverse model of each rate combination, there exists at least one equivalent forward rate combination. • Examples: • TIA-921: TC13 B2A same as A2B in 13,22,46&60 • TIA-921-A: TC13 B2A same as A2B in 49&73 • Some rate combos are symmetric and are thus their own reverse (e.g. 1,2,3) Anue Systems, Inc. www.anuesystems.com

  10. TIA-921 Reverse (B2A) Anue Systems, Inc. www.anuesystems.com

  11. TIA-921-A Reverse (B2A) Anue Systems, Inc. www.anuesystems.com

  12. Model behavior, CBR versus VBR • A fundamental assumption is that the model calculates new impairment values for approximately each packet • Thus we used 1ms update rate for 1k pps voice • And we used a 50us update rate for DTV • But, if traffic is bursty, and a whole burst of packets arrives during a timeslot which is marked for drop, the entire burst can get clobbered merely by being unlucky. • The model should have a way to handle bursty traffic more correctly. Anue Systems, Inc. www.anuesystems.com

  13. Time correlation of fwd/rev • For bidirectional emulation using the model, we must create two instances of the model • One for the forward link and one for the reverse. • These two models are not time correlated • Therefore congestion in the forward link is uncorrelated with congestion in the reverse link. Link failure and route flap are also uncorrelated. • I believe that in the real world there will be significant correlation between the two directions. • The model should have a way to mimic this behavior. • This leads to uncontrolled assymetry of delay • Certain services may be sensitive to delay assymetry • e.g. Two-way time transfer: NTP, PTP and CESoETH Anue Systems, Inc. www.anuesystems.com

  14. Conclusions • Bandwidth limiting and smoothing are important missing components of the existing model and should be added. • The rate combinations contain some redundancy that leads to confusion. Further work is required to determine if there’s a better way to designate the various rate combinations • VBR traffic can be “unlucky” and get zapped. Further works is required to determine how to address this • Forward and reverse models have no time correlation. This should be added to the model. Anue Systems, Inc. www.anuesystems.com

More Related