1. CAnet 4 Updates Thomas Tam June 11, 2006

2. Topics >L3 network update >L1 network update >L1 redesign >10GbE LANPHY testing >Connection program (CCP) – Damir

3. L3 network updates >Current Layer 3 network >Connections to the North >IPv6 multicast deployment >UNB and PEI backup peering >New BGP peers

4. Current Layer 3 network

5. Connections to the North – Physical details >Northwestel ATM circuit (over radio) from Whitehorse and Yellowknife to Ft. Nelson, and south to Ft. St. John, then to CAnet Edmonton PoP >Juniper M5 added in Edmonton POP (c4-edm01) – 4 port ATM DS3 Physical Interface card (PIC) – 4 port FastE PIC >c4-edm01 peers with c4-cal01 via STS-1 lightpath >Completion date – Yukon - June, 2005 – NWT - July, 2005

6. Connections to the North – circuit details

7. IPv6 Multicast >JUNOS bugs delayed deployment late summer. >Deployed late 2006. >Embedded-RP enabled on all routers. >Currently receive IPv6 multicast routes from Abilene, GEANT, AARnet, NYSERnet, BCnet and Fed. GigaPoP. >Not much testing over the last few months, looking for partners to test within Canada.

8. UNB and PEI backup peering >The idea was discussed for number of times but never found time to do the work. >The idea is that ORANs peer with each other and both announce same set of routes to CAnet 4. >In the normal situation, CAnet 4 routers select the routes from the directly connected ORAN due to the BGP longer AS path decision. >In mid February Blair, Ying and Damir were working together on routing policy to make it happened. >It is the first it’s kind on CAnet 4. >Since then multiple maintenance outages and circuit problems, backup peering maintained connectivity to CAnet 4.

9. New peers >National Lamba Rail (NLR) - AS19401 – A peer at Starlight, IPv4 and IPv6 peering to c4-win01 and c4- tor01 >LosNettos - AS226 – Los Angeles Regional Network, through CENIC infrastructure to Pacific Wave in Seattle. – IPv4 and IPv6 peering to c4-cal01

10. L1 network updates >HDXc deployment in Chicago and Seattle >4 th and 5 th waves >LP usage report >10G demos and activities

11. Current CAnet 4 network VanCal EdmSas RegWin Vic Sea TorOttMonHal Fre VanOttMonCal SeaChi KamKel NYC Chi Tor Win Edm CalSea CISCO ONS 1545Nortel OME 6500Nortel HDXc NYC Thu 1 st and 2 nd waves 3 rd wave 4 th wave 5 th wave

12. HDXc deployment >HDXc deployment in Chicago and Seattle – Both HDXcs were deployed in early Sept, 2005. – Seattle HDXc 8 x 10G circuits connected, 5 x CAnet 4 10G, KREONet2 10G, 10G to IEEAF(PNWGP) ONS and CENIC 10G 2 x 10G are planned in a month, IEEAF 10G and WIDE (U. of Tokyo) 10G – Chicago HDXc 9 x 10G circuits connected, 4 x CAnet 4 10G, 2 x 10G to Amsterdam, 3 x 10G to Starlight Force10. 2 x 10G are planned to connect OMNInet. >Both systems were used heavily during IGRID05 and SC05, were two of the major hubs in North American to provide connectivity to Asia and Europe.

13. 4 th and 5 th waves >Were delivered in October, 2005. >But three breakouts were competed in the later day – Kamloops, BC – Dec. 19 th, 2005 – Kelowna, BC – Nov. 30 th, 2005 – Thunder Bay, ON – May 12 th, 2006 >Both Kamloops and Kelowna breakouts allow BCnet to build their internal infrastructure. >The thunder Bay breakout provides a redundant path for the northern segment of the ORION network

14. Capacity allocation chart # 1

15. Capacity allocation chart # 2

16. Capacity allocation chart # 3

17. 10G demos and activities >McGill and UBC – Ultra Hi-Def Video conferencing project has taken up one of the 10G circuits. During SC05 they were putting 5Gbps of traffic from McGill to Seattle. >TRIUMF now connected at 10GbE, will be doing tests with CERN very soon once 10G circuit Amsterdam to CERN is ready. >All 10G circuits were fully utilized during IGRID05 and SC05. >U of Tokyo has been using one of our 10G circuits to participate in IPv4 and IPv6 Land speed record challenge.

18. Objectives – L1 redesign >meet CCP requirement >meet CAnet 4 scaling requirement >minimize soon-to-be obsolete ONS expenditures >minimize the overall costs

19. ONS 15454 - limitations >Maximum of 4 x OC192 cards per chassis >Maximum of 2 full speed GbE per cards, 8 slots chassis. Total of 16 full speed GbE ports >Few hub sites already or almost ran out of slot >No 10GbE LAN PHY support >Proprietary GbE encapsulation LAN Extension Protocol (LEX) to SONET framing, >GFP standard GbE encapsulation and VCAT are supported only on the new type of GE card, require hardware upgrade – new version of timing control cards (TCC) and GbE cards

20. Why VCAT? >SONET/SDH transport structures was designed to provide the traditional transport service (i.e. OC-3c/12c/48c/192c and more recently OC-24c) by combining multiple consective SONET channels. >VCAT was originally designed to extend the utility of the SONET/SDH transport layer. >VCAT provides a highly granular, flexible, and efficient method of provisioning traditional SONET/SDH transport bandwidth for packet-based data service transport. >VCAT allows for the grouping of non-consecutive SONET/SDH synchronous payload envelopes (SPEs) to create “virtual” concatenation bandwidth groups. >With VCAT, an arbitrary number (X) of virtual containers (STS-1/3c SPEs) are grouped together with the combined payload (STS-n-Xv) used to match the required bandwidth. >The combined payload isn’t limited going through one SONET path, can be through multiple SONET paths. >Here is some information on VCAT http://grid2.canarie.ca/wiki/index.php/Virtual_Concatenation_%28VCAT%29

21. Redesign criteria >Flexibility in routing LPs through different SONET paths >Allow GbEs to be dropped in either ONS and OME in the hub sites >Availability of slots for more 10G circuits if required. >GbE ports requirement for CCP and LPs >Maintain the continuity of manage infrastructure among different optical platforms >Minimize costs

22. Idea Scenario VanCal EdmSas Reg Win Vic Sea TorOttHal Fre VanOttMonCal SeaChi KamKel NYC Chi TorWin Mon NYC Edm CalSea CISCO ONS 1545Nortel OME 6500Nortel HDXc NYC Thu 1 st and 2 nd waves 3 rd wave 4 th wave 5 th wave

23. Idea Scenario >Breakout 4 th wave in Winnipeg – breakout cost >Re-termination of OC-192 circuits on OMEs – Vancouver, Winnipeg, Toronto, Montreal – 8 x OC192 cards >Migrate some of the LPs from ONS to OME – 7 x GbE cards >BUT, 8 ONS GbE cards are required for CCP needed

24. New topology >No more purchase of ONS cards >No breakout in Winnipeg, save on the breakout cost and cost of a OC192 card >Install a new OME in Victoria, Edmonton, and Saskatoon for CCP/LP needs >Redeploy one of the ONS in RISQ PoP, so LPs can be dropped on either ONS and OME. >Re-termination of OC-192 circuits on OMEs – Vancouver, Winnipeg, Montreal

25. Network Topology VanCal EdmSas RegWin Sea TorOtt Hal Fre VanOttMonCal Sea Chi KamKel NYCChi TorWin Mon #1 NYC Edm CalSea CISCO ONS 1545Nortel OME 6500Nortel HDXc NYC Vic Mon New OME Thu Mon #2 RISQ 1 st and 2 nd waves 3 rd wave 4 th wave 5 th wave

26. Migration schedule >Will be completed in the week of June 19 – June 19, 20 - OME install in Victoria – June 21, 22 – OME install in Edmonton, and Sasakatoon – June 22, 23 – ONE install in Montreal installation >At the same time, all CCP related work will be completed also.

27. Change on the connections to the North

28. 10GE LANPHY testing #1 >10GE LANPHY is a new type of the interface card for Nortel OME platform, the general availability is in the first quarter of 2007. The card will provide: – VCAT capacity, can be sub-rated down to STS3c – 2 x 10GbE LANPHY ports >The early deployment card is now available. CANARIE will purchase three cards for testing. Nortel will replace them once the general availability of the card is released. >It will support only – one GE LANPHY port and – sub-rated down to STS3c-32V (5G).

29. 10GbE LANPHY testing #2 >VCAT support give us the ability to utilize the SONET circuit efficiently. >For examples – McGill Ultra video conferencing – require only 5G but take up a whole 10G circuit – TRIUMF would have a 5G LP to CERN, multiple 1G to tier 2 site

30. 10GbE LANPHY testing #3

31. Connection Program (CCP) Damir Pobric