Академический Документы
Профессиональный Документы
Культура Документы
25
Contributing Co-Authors
z z z z z z z z z z z z z z z z
NTT: Eiji Oki Isocore: Rajiv Papneja Toyo Corp.: Shinichiro Morishita KDDI R&D Labs.: Kenichi Ogaki and Masanori Miyazawa Fujitsu Lab.: Keiji Miyazaki Fujitsu: Hiroaki Nakazato Juniper Networks: John Allen and Hidetsugu Sugiyama ITOCHU Techno-Solutions: Shinichi Hasegawa and Nobuhiro Sakuraba NEC: Itaru Nishioka Mitsubishi Electric: Shoichiro Seno OKI: Yoshihiro Nakahira Keio Univ.: Daisuke Ishii and Satoru Okamoto Agilent Technologies: Tara Van Unen Alcatel-Lucent: Mark Blumhardt Cisco: Hari Rakotoranto Sycamore: Vijay Pandian
Page 2
Outline
Status of Related Activities Study Objective Challenging Issues Experiments Results & Issues Remaining Problems & Discussion Conclusion
Page 3
IETF (Internet Engineering Task Force) z RSVP-TE (RFC 3473)/OSPF-TE (RFC4203)/LMP (RFC4204) z Protection & Restoration (RFC4872/RFC4873) z Current discussion is focused on
z Inter-domain signaling & routing in RFC editor queue z WSON z PBB-TE control z LCAS/VCAT control z ASON routing z Multi-layer (Nested domain architecture)
z MUPBED
Hans-Martin Foisel et al., ECOC2006 We3.P.119
JGN II
z NICT/NTT/KDDI/MCNC/Univ. of Illinois/Calient JOINT activities US-Japan Inter-carrier LSP control z Terminated four year field research activity
Page 5
SINET 3
SINET III constructed by National Institute of Informatics http://www.sinet.ad.jp/ Currently largest GMPLS network in Japan z 75 nodes includes STM256 (40 Gbit/s) links z Layer-1 BoD service z Users can specify destination, duration, bandwidth with a granularity of 150 Mbps, and route option. z BoD server receives reservation requests, schedules accepted reservations, and triggers Layer-1 path setup.
After hitlessly reducing bandwidths of L2/L3 paths by 1.8 Gbps using LCAS, established two L1 paths (0.9 Gbps x 2) on demand via client PC at Hokkaido Univ. Very stable transmission of non-compressed HDTV between sites
Hokkaido
BoD server
L1-OPS
NII (Tokyo)
L2 Mux Hokkaido
Client PC
Page 7
Objective
z Evaluate Intra-carrier multiple routing area architecture of IETF model - Per-area hop route calculation in area border (ABR)-OXC z Evaluate optical routing in STM-16/GbE multi-rate optical links z Evaluate routing in ROADM and OXC hybrid network
Page 8
Challenging Issues
Point II a) Auto-discovery of next-hop ABR-OXC b) Per-area hop route calculation at ABR-OXC c) Loose-hop expansion (RFC3209) at ABR-OXC
#A
Sub-area X Area 0
ABR-OXC ASBR-OXC
Sub-area Y
ABR-OXC
#Z
z Originate from scalability issue regarding open shortest path first protocol (OSPF) - Requires sub-area routing architecture if number of nodes in OSPF area exceeds about one hundred z OSPF-TE specification indicates that link Information is advertised within each routing sub-area - Link information is invisible to outside sub-areas
Page 9
Experimental Configuration
: ROADM : OXC
Area 2 @Toyo
#N2 #N1
ABR-OXC #L1
Area 3 @Toyo
Area 0 @Toyo
#K
#M1 #M2
#E1
#D3 #B1
MPLS NW
#C
#E2
#D2
#A2
#A1
Page 10
MPLS router
GMPLS routers
ROADMs TDM-XC
OXCs
Page 11
Path Creation
RSVP-TE:
z Switching Type: Lambda z Encoding Type: Ethernet or SDH z GPID: Ethernet or SDH
PATH
PATH
L
PATH PATH PATH
PATH RESV
#A
Sub-area X Area 0
ABR-OXC
Sub-area Y
ABR-OXC
*) PDCP: Per-domain path calculation
#Z
Page 12
RSVP-TE: Achieved very good inter-operability for strict ERO signaling OSPF-TE: Issues Encountered:
z Router LSA Vendor implementation does not support the advertisement of Node ID by stub area within Router LSA. This resulted in a lack of reachability information. What LSA or TLV should be responsible for providing reachability information to inter-domain LSPs ? Need discussion. z Router Address TLV ABR-OXCs did not advertise the Router Address TLV for sub-areas. Some vendor implementations recognized this as OSPF adjacency error and failed to perform CSPF calculation.
Page 13
Page 14
Switching constraint of ROADM z LSPs from West Tributary Ports are terminated at East Tributary Ports z This is also true for the opposite case. Ref. draft-imajuku-ccamp-rtg-constraint-0x.txt now merged to draft-bernstein-ccamp-wson-info-0x.txt
NNI link (West) West bound bi-directional LSP Add switches Drop switches ROADM Drop switches Add switches NNI link (East) East bound bi-directional LSP
RxTx
RxTx
RxTx
RxTx
Page 15
Issues Encountered:
z Auto-discovery of ABR-OXC/Next-hop ABR-OXC Auto-discovery of next-hop ABR-OXC is possible, if all GMPLS capable nodes advertise their node IDs within Router LSA. z Each ABR-OXC successfully performed per-area hop route calculation, if operators conduct manual assignment of next-hop ABR-OXC. ABR-OXCs automatically inserted explicit route objects into RSVP-TE messages (Loose hop expansion) to assign the route for their area.
PCE
#B2
#I
ABR-OXC #L2
#K
ABR-OXC #L1
30.204.16.13
Page 16
Three other successful scenarios for per-area hop route calculation in ABR-OXC
RTT 694 msec 5216 msec
#I #B1 ABR-OXC #L1 #K ABR-OXC #L2 ABR-OXC #L3 #B2
#N1
#N2
#I
583 msec
Per-area hop route calculation & loose hop expansion
Typically 8 to 9 seconds in addition to the round trip time is required to initiate IP packet forwarding.
Page 17
Results IV
RSVP PATH
58.92 sec
Page 19
Discussion
#A
Area 0
ABR-OXC ABR-OXC
#Z
Up to 20,000 reachability info, if UNI is provided for service network PCE IP reachability Info
Area 0
C-Plane ABR
#A
#Z
Page 20
Discussion (cont.)
Once we employ this control plane network architecture, sub-AS architecture is also applicable without changing deployed TNE control plane functionalities. Meaningful from the viewpoint of future proof.
OSPF Area 0
C-Plane ABR
#A
C-Plane sub-ASBR
#Z
BGP Sub-AS
PCE IP reachability Info
#A
#Z
Even in this case, IP reachability information is essential to discover next-hop PCE and ABR (or sub-ASBR), and to transport RSVP-TE notify messages between two end points.
Page 21
Conclusion
Achieved LSP creation with per-area hop route calculation in ROADM/OXC networks in four routing areas. Before addressing to inter-area protocol extension,
z Need to re-consider control plane network architecture z Need to consider how to locate PCE and ABR (or sub-ASBR). z need rough consensus for these issues. z While protocol extension has to have flexibility for the control plane network architecture, we have to find an effective way to avoid excess protocol extension for TNEs. z PCE architecture seems to be a good solution for this purpose, both for carrier and vendors.
Page 22