2. Tree Configuration in Bridged IEEE1394 Bus Network PHILIPS Research Subrata Banerjee PHILIPS Research Briarcliff P1394.1 WG Meeting March 19-20, 1998

3. Subrata Banerjee sub@philabs.research.philips.com PHILIPS Problem Statement Bridges of different capabilities  Bandwidth  Iso Delay Bridges may introduce unacceptable loops Path between any two bridges  “Minimum bottleneck” route Typically bridge capacity lower than bus capacity N1 C1C2N2

4. Subrata Banerjee sub@philabs.research.philips.com PHILIPS Graph Theory Result Maximum Spanning Tree guarantees Minimum Bottleneck Route Between every pair of nodes Distributed maximum spanning tree algorithm C2 C1

5. Subrata Banerjee sub@philabs.research.philips.com PHILIPS How to Choose Between Two Bridges? Bridge Capability Parameters  Bridge Bandwidth  Bridge Iso_Delay  Bridge Vendor ID = max. of 2 portal vendor IDs  Bridge Node ID = max. of 2 portal vendor IDs Proposed Organization Bridge BW (13) Resv (3) Iso_delay (8)Resv (8) Chip ID (40) Vendor ID (24) (scrambled)

6. Subrata Banerjee sub@philabs.research.philips.com PHILIPS An Example of Tree Conf. Algorithm (1) 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36

7. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 BP62 An Example of Tree Conf. Algorithm (2) BP Broadcasts BP31 BP40BP52 BP65 BP27 BP36BP38 BP48 BP21 BP57BP55 BP58 BP34

8. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 BP Count=2 An Example of Tree Conf. Algorithm (3) Fragment Roots BP Count=3 BP Count=2 BP Count=4 62, L1 58, L1 57, L1 48, L1 65, L1

9. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (4) Find Best Neighbor 62, L1 58, L1 57, L1 48, L1 65, L1 Submit

10. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (5) Join 1 62, L1 58, L1 57, L1 48, L1 65, L1

11. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (6) Join 1 58, L1 57, L1 48, L1 65, L2

12. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (7) Find new best neighbor 58, L1 57, L1 48, L1 65, L2

13. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (8) Join 2 58, L1 57, L1 48, L1 65, L2

14. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (9) Join 2 58, L2 48, L1 65, L2

15. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (10) Join 3 48, L1 65, L2 58, L2

16. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 38 36 An Example of Tree Conf. Algorithm (11) Join 3 58, L2 65, L3

17. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 36 An Example of Tree Conf. Algorithm (12) Join 3 58, L2 65, L3

18. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 27 48 57 36 An Example of Tree Conf. Algorithm (13) Join 4 58, L2

19. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 31 38 48 40 27 21 36 34 58 55 65 58 48 57 An Example of Tree Conf. Algorithm (14) Join 4

20. Subrata Banerjee sub@philabs.research.philips.com PHILIPS 62 65 52 57 48 40 58 55 65 58 48 57 An Example of Tree Conf. Algorithm (15) Join 4

21. Subrata Banerjee sub@philabs.research.philips.com PHILIPS Rules of the Game Fragment A can submit to Fragment B iff Level A  Level B Bridge Capability values are unique. No loops possible LxLy Lz Lx > Lz > Ly > Lx Lx C3 C2 C1 C1 > C2 > C3 > C1

22. Subrata Banerjee sub@philabs.research.philips.com PHILIPS Example of Selected Commands “I am a BP”  data = Unique Bridge Capabilities (UBC) “Report UBC”  data = best UBC from all children “Connect”  data = Fragment Level “Update”  data = Fragment UID, Fragment Level

23. Subrata Banerjee sub@philabs.research.philips.com PHILIPS Once the Tree Topology is Identified... Assign  Bus Ids  Routing Bounds  Portal_Control.rte fields Select Net Cycle Timer Assign Portal_Control.clk fields Details?