Presented by Elisa Belllagamba,
Portfolio Strategy Manager, Business Unit Networks, Product Area IP & Broadband, at the MPLS and Ethernet World Congress 2012
Virtual Network System: Splitting Control and Data Plane to Optimize IP and Transport Services at Minimum OPEX
1. Virtual Network System:
Splitting Control and Data Plane
to Optimize IP and Transport Services at Minimum OPEX
Elisa Bellagamba
elisa.bellagamba@ericsson.com
2. Trends & driving forces
Internet bandwidth growth
Mobile Traffic Growth
Personal, mobile cloud services
50B Connections
3. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
MOBILE
CORE
Revenue =
MOBILE
ƒn (competition) +
ƒn (content ownership)
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED
MULTI-SERVICE
EDGE
DEVICES ACCESS AGGREGATION IP EDGE CORE
4. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
How to extend IP services
up to the access? MOBILE
CORE
Revenue =
MOBILE
ƒn (competition) +
ƒn (content ownership)
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED
MULTI-SERVICE
EDGE
DEVICES ACCESS AGGREGATION IP EDGE CORE
5. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
How to extend IP services
up to the access? MOBILE
CORE
Revenue =
MOBILE
Upgrade ƒn (competition) +
boxes to ƒn (content ownership)
full IP?
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED
MULTI-SERVICE
EDGE
DEVICES ACCESS AGGREGATION IP EDGE CORE
6. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
How to extend IP services
up to the access? whole
Train the
MOBILE
CORE
Revenue =
personnel to
ƒn (competition) +
MOBILE
TDM TDM build IP ƒn (content ownership)
competence?
TDM TDM
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
TDM
TDM TDM TDM ƒn (traffic growth)
MUST BREAK THIS!
FIXED
TDM TDM TDM
MULTI-SERVICE
EDGE
TDM TDM TDM TDM
DEVICES ACCESS AGGREGATION IP EDGE CORE
7. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
How to extend IP services
up to the access? whole
Train the
MOBILE
CORE
Revenue =
personnel to
IP IP ƒn (competition) +
MOBILE
build IP ƒn (content ownership)
IP IP competence?
BACKHAUL & METRO TRANSPORT & ROUTING
IP IP IP IP Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED IP IP IP
MULTI-SERVICE
EDGE
IP IP IP IP
DEVICES ACCESS AGGREGATION IP EDGE CORE
8. Control Plane options
for access and aggregation networks
Today’s TDM NetworksConverged Network Today’s IP Networks
NMS NMS NMS
Centralized
control plane
Static Dynamic Dynamic
Centralized control Centralized Control Distributed control
Simplified operational model IP operational model
Data Plane Control Plane Interface to Common Control Plane
Aggregation Core Aggregation
9. Traditional networking:
functional representation
traditional traditional
ROUTER TRANSPORT
NODE
NMS
NMS, PCE
Mgmt, CLI, SNMP Svc cfg TED
RIB FIB LIB LFIB Mgmt, CLI
XC dB*
Signaling/TE (RSVP, LDP, CSPF)
IGP (for top. discovery, DCN)
Routing (IGP, BGP)
Forwarding, OAM handling
Forwarding, OAM handling
Control plane is baked into * For MPLS-TP,
underlying forwarding can be
infrastructure, that poses considered LFIB
scale, and feature velocity
problems
11. Virtual Network System:
the network as a platform
ACCESS AGGREGATION EDGE CORE
Virtual Network System
CSR IP
TCN
RESIDENTIAL
PE
MTU
ENTERPRISE
CPE
Virtual Network System can logically represent as a single entity
a full cluster/region of nodes
12. CP consolidation in Aggregation
› Homogenous DP: operational simplicity › The centralized control plane element
› CP consolidation: topology hiding, can act as a single IP/MPLS node
manage scale participating in the IP core, on behalf of
the NEs in the access/aggregation
INDEPENDENT SCALING OF CONTROL PLANE AND DATA PLANE
IP/MPLS Metro
Control plane
i/eGP Domain
Data plane
13. TCO analysis
1ST 2ND 3RD
Access Aggregation Aggregation Aggregation
VNS
domain
~70%
OPEX
SPO1410 SSR
GE
… GE TCN savings
10GE
SPO1410
TCN
Download the paper here
14. Connectivity Scenarios details:
p2p L2 services
ACCESS METRO EDGE CORE METRO ACCESS
1st Aggr. 2nd Aggr.
VNS
domain
AN VNS
AG1 AG2
PE
ABR
AG1
AN AG2
CE PE PE PE
Eth AC LDP PW
Fixed/Mobile Backhaul
Business VPN
Business VPN, backhaul (e.g. X2)
15. Connectivity Scenarios details
mp2mp L3 services
ACCESS METRO EDGE CORE METRO ACCESS
1st Aggr. 2nd Aggr.
VNS
domain
AN VNS
AG1 AG2
PE
ABR
AG1
AN AG2
CE PE PE CE
Business VPN
Business VPN, backhaul (e.g. X2)
16. Connectivity Scenarios details:
mp2mp L3 services
ACCESS METRO EDGE CORE METRO ACCESS
1st Aggr. 2nd Aggr.
VNS
domain
AN VNS
AG1 AG2
PE
ABR
AG1
AN AG2
CE PE PE CE PE CE
Eth AC LDP, iBGP
17. Running demo in Ericsson booth
Show
VNS domain L3VPN
R3 10.50.3.1 GUI &
10.33.99.3 vrf1 CMD Line
video
TCN
client
DPN 2
SPO-1410
Static L3
laptop 1 VPN
10.50.3.2 OSPF, configuration
BGP,
LDP
video
DPN 3
ML-SP 210 10.50.1.1 server
SPO-1410
Plug-in new DPN 1 vrf1
ML-SP 210
SE100 SE100 laptop 2
Auto- 10.50.1.2
Provisioned R1 R2
ML-SP 210 10.33.99.1 10.33.99.2
DPN 3
Adding a “customer site”, auto-configuration of VNS internal transport
Configure & prove L3VPN by TCN command line printouts
18. Testing the controller scalability:
study assumptions
CONNECTIVITY
1 MODELS Several combination
MSER GGSN
PE
IPTV Internet
PGW server
of mesh, ring and
tree topologies with
AGS1
up to 3 level of AGS1
AGS2
aggregation AGS2
FIBER AGS1
ACCESS AGS1
PE PE
ABR ABR
FIBER
ACCESS
AGS2
FIBER
AGS2 ACCESS
AGS1
AGS1
FIBER
ACCESS
FIBER
VNS domain ACCESS
Quantitative
2 SERVICES 3 assumptions
› MBH › Static › Dynamic
– number of – Link down
› IPTV Tunnels › affected tunnels
– number of › timescale
› p2p Equipments – IPTV channel change
› timescale
19. Testing the controller scalability:
RESULTS
CONNECTIVITY
1 MODELS
Which are the requirement on the VNS
controller in order to enable the
controlled network to have same or
2 SERVICES better performance than a
traditional distributed network while
providing additional functionality
Quantitative
3 assumptions
and reduced switch complexity?
› Static behavior
– Requirements are in the order of magnitude or below the capabilities of
existing controllers and switches
› Dynamic behavior (restoration)
– A centralized restoration scheme coupled to a data plane protection meet the
transport requirements of 50ms while ensuring best path availability
– Changing the connection structures has a significant impact on the scalability
20. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
MOBILE
CORE
Multilayer Revenue =
optimization?
MOBILE
ƒn (competition) +
ƒn (content ownership)
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED
MULTI-SERVICE
EDGE
DEVICES ACCESS AGGREGATION IP EDGE CORE
21. Leveraging on multilayer technology
application
aware routing PKT PKT
PKT
PKT
OTN
OTN
OTN OTN
variable WDM
bandwidth WDM WDM
WDM
packet links
unified
recovery
There is an opportunity to improve the coordination between Packet
and Optical layers with a fully dynamic coordination
22. Rapid change carries a cost
Capacity =
Dealing with Scale Lowering Costs Creating Revenue ƒn (demand) +
ƒn (performance)
MOBILE
CORE
Revenue =
MOBILE
ƒn (competition) +
ƒn (content ownership)
BACKHAUL & METRO TRANSPORT & ROUTING
Network costs =
ƒn (traffic growth)
MUST BREAK THIS!
FIXED
MULTI-SERVICE
EDGE
DEVICES ACCESS AGGREGATION IP EDGE CORE
23. Premium OTT content delivery and
Inline Service Chaining
EDGE
GGSN PCRF
MSER PGW
CSR
VNS
PE Inline service
ABR
chaining,
VNS traffic
domain steering
CPE
OTT Player makes use of network capabilities to
enhance its services for the users
24. Conclusions
MAINSTREAM: IP implemented in every node
Growing complexity
NEW PARADIGM: Virtual Network System
Simplified management
Supporting growth and service variety