Corrigent ConfidentialCopyright © 2007 Corrigent Systems
100G Packet Ring ArchitecturesGady RosenfeldVP Marketing [email protected]
October 2007
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
2
The need for 100G
Cox – "100GE needed for broadband customer aggregation urgently in the core by 2009 and across the board by 2011", John Weil, Apr'07
Comcast – “There is a market need for 100GE”, Vik Saxena, Jan’07
Equinix – Requirements for “100 Gbps or greater”, Louis Lee, Jan’07
Level 3 – Using 8x10 GbE LAG today
Yahoo! – Using 4x10 GbE LAG today
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
5
Generic Triple-Play Network Architecture
National Video Content
Distribution Network(IP Multicast)
Local IPTV Video Distribution Network
NHO
BRAS
Regional Content Insertion
National Content Insertion
NHO Metro Node
ISP1
ISP2
ISP3
Metro Node
DigitalVideoServer
VoIP
VoIPDigitalVideoServer
IP Core Network(Tier 1 aggregation network)
Metro Transport Network (Tier 2 aggregation network)
Local Distribution
Local IPTV Video Distribution Network
MGW
VideoAcquisition
System
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
6
Triple-Play Network – Metro Transport
ER : Edge Router (Layer-3)
PA : Packet Aggregator (Layer-2) FiberCopper
Customer Premises Packet transport switch
Local Content Insertion
Metro Node
DigitalVideoServer
ER
DSLAM
PA
Nx10G
Metro Node
DigitalVideoServer
ER
10G metro rings
PA
PA
KEY
Nx10GE
Nx10GE
PTS
PTS
PTS
PTS
PTS
PTS
PTS
PTS
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
7
Bandwidth Requirements
IPTV
• 2007 – 300 channels, 10% HD: 1.1-1.4 Gbits/s (MPEG-4/MPEG-2)
• 2010 – 300 channels, 50% HD: 2.0-3.2 Gbits/s
VoD (2500 subscribers per node)
• 2007 – 5% VoD penetration: 0.5-0.6 Gbits/s
• 2010 – 30% VoD penetration: 5.0-8.0 Gbits/s (MPEG-4/MPEG-2)
Total bandwidth requirements – 6 nodes per ring
• 2007 – 3.5-4.5 Gbits/s
• 2010 – 32-51 Gbits/s
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
8
IEEE 802.3 HSSG Status
IEEE 802.3 HSSG
• Agreed on PAR for 40GE and 100GE, July’07
• Identify bandwidth-hungry applications: data centers, internet exchanges, high-performance computing and video on demand
Parallel optics for 100GE (4x25G, 10x10G) discussed for dedicated fiber and limited distances applications. Serial options for MAN/WAN applications still under evaluation
• Polarization multiplexing, Phase coding
Standard is still at least 4 years away
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
9
Alternative for Network Scalability
Add separate rings
• Complex network operation – multiple networks, Traffic Engineering
• No redundancy between rings
• Limited statistical multiplexing
Upgrade to 40 Gbits/s
• Disruptive and costly process
• High equipment cost – optics, network processors, traffic management
• Limited capacity
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
10
High-Capacity Packet Rings
PTS PTS
PTS
PTS
PTS PTS
PTS
PTS100G MAC Layer
nx10G PHY
High-capacity (HC) packet rings are achieved through advanced bonding techniques
Multiple 10G RPR instances are combined to create a single logical ring
40G links can also be added to the bundle
Flow-aware hashing for load balancing and distributing packets over parallel physical links
Guarantees traffic integrity, by uniquely identifying and classifying each individual flow over the same physical link, avoiding re-ordering
RPR#2
RPR#1
Hashing
UserInterface
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
11
HC Packet Rings – Traffic Distribution
No mis-ordering within a flow
• Each flow is consistently delivered on the same channel
• Packet ordering is maintained even if each channel is carried in different route with different length
• Flexible combinations of fields used for hashing to provide load balancing in different applications
1234
1234
1234
1234
1234
1234
1
1
1
1
1
12
Transmitted packets over 4 channels
6 flows
23
2
2
3
Link Failure
After the failure packets are distributed over 3 channels
2
4 4
2
3
3
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
12
HC Packet Rings Survivability
TDM Flow
DataFlow
RPR Steer protection
- Logical port
- Physical RPR MAC
RPR Steer protection
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
13
HC Packet Rings Enhanced Survivability
- Logical port
- Physical RPR MAC
TDM Flow
Dataservice
RPR Link#2 is Down
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
14
Customer ACustomer A
Customer ACustomer A
Customer ACustomer A
Customer BCustomer B
Customer BCustomer B
Customer BCustomer B
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer BCustomer B
L2-VPN service to interconnect between enterprise's branches
VPLS over ring network
Can be infrastructure service to multiple end-user services
Network CapacityCustomer A – 3GCustomer B – 3GCustomer C – 4G
Total net capacity : 10G
Network CapacityCustomer A – 3GCustomer B – 4GCustomer C – 4G
Total net capacity : 11G
Example – Growth of Existing Services (1/3)
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
15
Customer ACustomer A
Customer ACustomer A
Customer ACustomer A
Customer BCustomer B
Customer BCustomer B
Customer BCustomer B
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer BCustomer B
Option 1 – Multi-ring configuration
Add additional ring instance – ringlet #2
Disconnect all CustomerB locations from ringlet #1
Re-provisioning Customer B service on ringlet #2
Example – Growth of Existing Services (2/3)
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
16
Customer ACustomer A
Customer ACustomer A
Customer ACustomer A
Customer BCustomer B
Customer BCustomer B
Customer BCustomer B
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer CCustomer C
Customer BCustomer B
Option 2 – HC-RPR
Increase RPR ring capacity to 20G
Connect Customer B 4th location to the existing L2-VPN service
Example – Growth of Existing Services (3/3)
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
17
Multi-Phy HC Packet Rings
Description • Allow combination of RPRoSTM64 and RPRo10GE in the same HC-RPR
group.
Motivation• Reduce cost while maintaining ring synchronization. • Clock distribution across the ring via SONET/SDH interface • Data and TDM traffic will run on top of both Ethernet and SONET/SDH
interfaces – full flexibility
Implementation aspects • Eliminate miss-order by per flow hashing • Fine flow granularity to assure equal load sharing between RPR instances
• Flow granularity: MAC ( S+D) + IP (S+D) + Port
• No issue of equal load sharing between different Phy layers • OC192 payload rate (net rate): 9.51Gbps• 10GE tri-model average payload rate: 9.5Gpbs
Equal net rates
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
18
Asymmetric Operation (AHC-RPR) and Management
Best for incremental network growth
Install RPR blades and optics only as node capacity demand increases
At least one ring must be common to all stations
Each station is represented by HC (group) MAC and physical MAC
• HC MAC is used for data forwarding and IP level
• Physical MAC used for topology
Reference topology has group entity and per ring entities
2x10GHC-RPR
PTS
PTS
PTS
PTS
PTSS1
S2 S3
S4
S5
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
19
The CM4000 Packet Transport Switch Layer
Transport PlaneM
on
ito
rin
g,
Su
rviv
abil
ity
and
mu
ltip
lexi
ng
SONET/SDH Ethernet
SONET/SDH Line
SONET/SDH Path
1 GE RPR10 GE Nx10GE
Classification Marking Queuing Tagging Policing
Multipoint
Interworking
Point to point Point to Multipoint
Ethernet IP/MPLS PPP FCTDM HDLC
Packet-based Path/Link Technologies Packet-based Multiplexing, Survivability and Monitoring at the
Path/Link layers
OTN (G.709)
MPLS LSP
NxRPR
Corrigent ConfidentialCopyright © 2007 Corrigent Systems
20
Summary
HC Packet Transport
• Network scalability up to 100 Gbits/s for high bandwidth applications is required today
• 100GE is at least 4 years away
• Cost effective network migration path is required
• In-service network scalability in 10G or 40G increments
• Resiliency to fiber and equipment failures
• Implemented with available low-cost optical components
Top Related