2nd Workshop on Future Internet Design
Evolution of virtualization
Evolution of virtualization
technology for the future optical
internetworking
Michiaki Hayashi
[email protected]
KDDI R&D Laboratories Inc
KDDI R&D Laboratories Inc.,
2-1-15 Ohara, Fujimino-shi, Saitama 356-8502, Japan
+81 492 78 7547 +81 492 78 7821
Agenda
Networking evolutions
Network resource virtualization for telecom operator
Network resource virtualization for telecom operator
High-level architecture
Design of prototypes
Design of prototypes
Demonstration
Policy-based E2E quality control
y
q
y
Resource scheduling
Distributed messaging flow
Virtualization of functional modules
High-level architecture
C
h f
ti
t l
Cache function control
Conclusion
Sep. 21st, ‘08
2 2nd Workshop on FID
Networking evolutions
From consolidation to customization
-From consolidation to
customization-~Today
“NGN”
Beyond NGN
e
y
o
n
n
Virtual network
ADM ADMIP/Ethernet
a
nsport
h
itectur
e
s
olidati
o
Storage(e g cache)
o
mizatio
n
TDM
Tr
a
arc
h
IP/Ethernet
Con
s
Data processing
(
i
)
Storage(e.g. cache)
Cust
o
ATM
n
g
ies
Control plane
Session/flow control
(e.g. compression)
Virtualization
N
etworki
n
c
hnolog
(MPLS, GMPLS)
Control plane
(MPLS GMPLS)
(IMS, RACF, BGF)
(Network resource,
Functional
modules)
Sep. 21st, ‘08 2nd Workshop on FID 3
Virtualization: Advantages
Short delivery time
Prompt coordination of heterogeneous infrastructures
p
g
Low cost
Good reusability and high utilization of infrastructures
Flexibility
Customized network and other building blocks
Connectivity
Customized network
Agenda
Networking evolutions
Network resource virtualization for telecom operator
Network resource virtualization for telecom operator
High-level architecture
Design of prototypes
Design of prototypes
Demonstration
Policy-based E2E quality control
y
q
y
Resource scheduling
Distributed messaging flow
Virtualization of functional modules
High-level architecture
C
h f
ti
t l
Cache function control
Conclusion
Sep. 21st, ‘08
5 2nd Workshop on FID
Virtualization: Grid approach
5
2Gbps
Site A 5
From G-lambda project (www.g-lambda.net)
Grid Application
Grid Portal
1 2Gbps 10 Site B Site C Reserved time : hhmmss hhmmssApplications
1 2Gbps 10 Duration : x minGrid Resource Scheduler
hhmmss - hhmmss
Result
Deadline : hhmmssRequirement
Computing Resource Managers
Network Resource Managers
Middleware
for virtualization
#6#
6
#6#6
# #
Computers
Resource/
Fabric
Sep. 21st, ‘08 2nd Workshop on FID 6
#6#6
#6#6
Seeds of virtualization in NGN
From ITU-T Rec. Y.2012
IMS
(IP Multimedia Subsystem)
RACF: Resource
virtualization for IMS
Sep. 21st, ‘08 2nd Workshop on FID 7
Network resource virtualization
Based on Grid approach
- Based on Grid
approach-
Virtualization
: NRM as a virtualization engine for various resources
Common API
: Network service interface (NSI) for coherent accessibility
Common API
: Network service interface (NSI) for coherent accessibility
Business process
: Harmonization to the business processes
Application
Provisioning system
Other OSS
SOA b
Network service clients
Common API
over SOA bus
Business
Process
Oriented
SOA bus
NRM
NRM
NRM
NRM
NRM
NRM
- NRM: Network Resource Manager SOA S iVirtualization
Engine
Network
1
Network
2-1
Network
2-2
Network
3
- SOA: Service Oriented ArchitectureEngine
Sep. 21st, ‘08 2nd Workshop on FID 8
1
Network 2
Network as a service in SOA
Loosely coupled
Tolerant to any change of backend systems (Good reusability)
Implementation environment
Rich frameworks and development tools (Lower development cost)
Business driven work flow
Easy integration to the existing business process
Accounting
Contents
Backend
systems
SOA
supply chain
User profile
Contents
P
i i
i
Inventory
Network
Provisioning
Inventory
Alarm
Metering
Virtualization of optical networks
R
t
iddl
d b KDDI R&D l b
Resource management middleware proposed by KDDI R&D labs
NRB:
Single point of contact
to service control layer
NRMs:
Distributed
management to
cover E2E
Services control
Alarm Application
SDP Configuration system
Network service clients
IMS/MMD NRB: Network Resource BrokerNRM: Network Resource Manager
layer servers Inventory Traffic
NRB
Orchestration
Resource management system A-NRM M-NRM C-NRM A-NRM SOA-busVirtualization
layer
Orchestration
layer
management system Virtualization Orchestration Data center EMS Access Access Resource control layer Metro Metro s port layer VoIP P2P CPE IPTV OLT Core (Mesh) VoIP P2P IPTV CPE OLT Si Si SiSi (Ring) (Ring) Tr a n s P2P Video-conference Router P2P Video-conference Si Si Si Si SiSi Si Si Si Si SiSiDesign of virtualization engines
Hierarchical
path computation with NRB-NRM
load sharing
resource
management architecture
NRB:
Abstracted
topology handling and
p
gy
g
parallel transaction
p
handling
g
NRM:
Detailed
topology handling, resource
scheduling
and deciding
resource allocation
policy
Universal WSI*
for reconfigurable extensibility
NRB
Network service clients
NRM
NRB
*WSI: Web Services Interface Universal WSI
NRB Web services module
Web AS WSRF
Web services module
Web AS WSRF NRM Transaction Transaction DB Mediation module Mediation module Path comp.
Parallelizing Roll back Abstracted Path comp. Scheduler Transaction
DB
Policy DB Resource DB Web services module
Web AS WSRF
g
topology DB
Network C&M module
Resource DB
Web AS WSRF
NRMs
Network C&M module
NE or EMS
Sep. 21st, ‘08
11 2nd Workshop on FID
GE-PON virtualization
M
t
li i
d
h
i
Management policies and mechanisms
Policy-based
CIR/PIR
provisioning for
each LLID (<4 per user)
Time-scheduled
bandwidth management of
g
shared PON link
Call admission control
and
degeneration
management
On-demand
and
scheduled
services
DBA*
for unused bandwidth
*DBA: Dynamic Bandwidth AllocationDBA: Dynamic Bandwidth Allocation
OLT ONU/HGW IPTV 4 Shared PON link Service path segment A-NRM e d-CIR
Admission control line (1Gbit/s)
LLID #1 LLID #2 LLID #3 1 Gbit/s link VoIP P2P/ data IPTV 4 6 0 T otal reserv e LLID #4 ToS
markerClas-sifier
4 1 Gbit/s max. 1 Gbit/s link from/to Metro ring data Video-conference 5 Policer o S marker Σ Time T LLID#1 LLID #253 LLID #254 LLID #255 4 6 0 s sifier C o QoS CIR (%) PIR (%) 0 100 100 Policy table R assigned LLID#1 LLID#2 LLID#3 LLID#4 LLID #256 ONU/HGW 5 Cla s 0 100 100 7 0 100 Time CI R LLID#4 LLID#N Sep. 21st, ‘08 12 2nd Workshop on FID
RPR virtualization
Management policies and mechanisms
Policy-based CIR/EIR provisioning for each VPLS path (<1023 total)
Cl
ifi
i
f
i
i h
VLAN ID
d Q S
Classification of services with
VLAN-ID and QoS
Time-scheduled
bandwidth management pf shared RPR link
Call admission control
and
degeneration
g
management
g
On-demand
and
scheduled
services
2.4 G link from/to a neighbor
Shared RPR link Service path
Admission control line (2.4Gbit/s) e d BW RPR node Service path (VPLS path) HP QoS CIR (%) EIR (%) otal reserv e BW assignment Policy tables from /to OLT HP G1 G2 (%) (%) 0 100 0 7 0 100 Σ Time T o O b ki t VPLS#1(HP)
Clas- Poli
Clas-BE Overbooking rate Class Rate (%) HP NA G1 200 VPLS#2(G1, G2) ( ) V PLS BWs sifier 2.4 G link from/to a neighbor -cer sifier G1G2 1,000200 BE 10,000 Time VPLS#N(BE) V Sep. 21st, ‘08 13 2nd Workshop on FID
IP/lambda virtualization
Management policy and mechanisms
Management policy and mechanisms
Request-based adaptive
LSP allocation
In-advance path computation
with breadth-first search
LSP
l
ti
ti
l t
d b
d idth
i
t
LSP selection meeting
latency
and bandwidth requirements
Time-scheduled
bandwidth management with
per-link basis
Admission
control and
degeneration
management
O d
d
d
h d l d
i
On-demand
and
scheduled
services
1Gbit/s 2.4Gbit/s
3
rdDifferent
li k tt ib t
Link ID Resource scheduling table per link 2 Gbit/ 2 4 Gbit/
2G path (A-B-C-D-E) 2.4G path (A-B-I-J)
LSC PSC
5ms 6ms 1ms
1
st2
nd3
rd
link attributes
A-B#1 B-C#1 C-D#1 2 Gbit/s 2 Gbit/s 2 Gbit/s 2.4 Gbit/s A 1ms B C D E 4ms 6ms 7ms 3msLoop
D-E#1 B-I#1 J-I#1 2 Gbit/s 2.4 Gbit/s 2.4 Gbit/s 1ms 7msp
detection
J-I#2 I-G#2 G-F#2 500 Mbit/s 500 Mbit/s 500 Mbit/s J I G F 1ms 8ms G F#2 ・・・ Time 500M path (J-I-G-F) Sep. 21st, ‘08 14 2nd Workshop on FIDAgenda
Networking evolutions
Network resource virtualization for telecom operator
Network resource virtualization for telecom operator
High-level architecture
Design of prototypes
Design of prototypes
Demonstration
Policy-base E2E quality control
y
q
y
Resource scheduling
Distributed messaging flow
Virtualization of functional modules
High-level architecture
C
h f
ti
t l
Cache function control
Conclusion
Sep. 21st, ‘08
15 2nd Workshop on FID
Demonstration
Southbound implementation of NRMs
A-NRM: CLIs of OLT, C-NRM: CLI of router, M-NRM: CORBA of EMS
Typical server platforms for NRB and NRMs
Typical server platforms for NRB and NRMs
Memory: 768 Mbytes
CPU: 2.4 GHz
G-NRM M-NRM1 L2 switch A-NRM Client NRB M-NRM2 CPE 1 CLI (of OLT) EMS CORBA CLI(of edge router) EP 1
RPR2 Si Si C Si Si Si Si OLT1 Si Si Si Si Si Si Metro Metro CPE 2 CPE 3 EP 2 RPR1 RPR6 EP 4 EP 5 Si Si SiSi Si Si Si Si SiSi Si Si C (Ring 1) (Ring 2) C t l l CPE 3 EP 3 RPR3 EP 6 RPR5 RPR7 EP 5 Core Control plane 1ms 3ms 5ms 10ms 1ms Rtr1 XC1 XC2 Rtr2 2.4Gbit/s OLT2 RPR4 RPR8 Demonstration environment 3ms 1ms 10ms 1ms 7ms Rtr3 XC3 XC4 Rtr4 1Gbit/s 600Mbit/s
Policy-based E2E quality control
Pre-planned (DiffServe-based)
Entire services
are
degraded
without
admission control and BW management
Traffic Typical
service BW profile
g
Degradation of
entire video services
ToS=6 Voice 90kbps*(200 flows)ToS=5 Real-time video
6Mbps*(N flows)
NRM/NRB-controlled
E2E CIR
i
d
d
d
videoToS=0 Data 5Mbps*(140 flows)
NRM/NRB-controlled
Pre-planned (DiffServe)
E2E CIR
is assured
on-demand
Voice
and
video
services are
preserved
Sep. 21st, ‘08
17 2nd Workshop on FID
Resource scheduling
S
f l
hi
hi
l
th
t ti
d
h d li
Successful
hierarchical path computation and scheduling
Detailed
route adaptation
by C-NRM (Request 3)
Abstracted rerouting by
roll back
operation of NRB (Request 4)
Request # EP 1 EP 2 Bandwidth (M bit/s) Latency (ms)Scheduled time frame
1 Rtr1 Rtr2 500 8 11:00 10:00 12:00 13:00 14:00 15:00 16:00 17:00 2 CPE1 EP1 300 10 3 CPE2 EP2 300 22 4 CPE3 EP3 350 10 5 CPE3 EP1 300 7
Request # Initial route selected by NRB Final assigned route (w/ assist of NRMs)
1 Rtr1-Rtr2 Rtr1-XC1-XC2-Rtr2 2 CPE1-OLT1-RPR1-RPR4-Rtr1-Rtr2-RPR8-RPR6 CPE1-OLT1-RPR1-RPR4-Rtr1-XC1-XC2-Rtr2-RPR8-RPR6 3 CPE2-OLT1-RPR1-RPR4-Rtr1-Rtr2-RPR8-RPR7 CPE2-OLT1-RPR1-RPR4- Rtr1-XC1-XC3-XC4-XC2-Rtr2-RPR8-RPR7 4 CPE3-OLT1-RPR1-RPR4-Rtr1-Rtr2-RPR8-RPR7 CPE3-OLT1-RPR1-RPR4-Rtr3-XC3-XC4-Rtr4 -RPR8-RPR7 RPR8 RPR7 RPR8 RPR7 5 CPE3-OLT1-RPR1-RPR4-Rtr1-Rtr2-RPR8-RPR6 CPE3-OLT1-RPR1-RPR4-Rtr1-XC1-XC2-Rtr2-RPR8-RPR6
Distributed messaging flow
Successful roll back
with
two-phase commit
(request #4)
Client
NRB
A-NRM
M-NRM1
C-NRM
M-NRM2
Reserve/Abort Reserve Reserve/Abort Roll-back in action Reserve/Prepared Reserve/Prepared Status Query Initial h ase Reserve/Abort Reserve/Abort Reserve/Prepared Reserve/Prepared Status Query Prepared 1 st p h Commit Commit Commit Commit phase Commit nd 2 Status Query CommittedAgenda
Networking evolutions
Network resource virtualization for telecom operator
Network resource virtualization for telecom operator
High-level architecture
Design of prototypes
Design of prototypes
Demonstration
Policy-based E2E quality control
y
q
y
Resource scheduling
Distributed messaging flow
Virtualization of functional modules
High-level architecture
C
h f
ti
t l
Cache function control
Conclusion
Sep. 21st, ‘08
20 2nd Workshop on FID
Virtualization of functional modules
Functional modules building blocks
Functional modules building blocks
Load
balancing Encap Compress DPI Protocol Redundancy Firewall
security memory
On-demand/work-flow
FC E
VPN
On-demand/work-flow
FCoE + VPN
CIFS optimization + Compress
Secured FCoE
WAN boost
p
p
Load balancing + SBC
VoIP media GW
Scalable
L4-SW
Sep. 21st, ‘08
21 2nd Workshop on FID
Virtualization of functional modules
NRB N t k R B k Inventory Traffic Alarm Application servers SDP Configuration systemNetwork service clients
IMS NRB: Network Resource BrokerNRM: Network Resource Manager Function Manager
NRB management systemNetwork resource
Orchestration NRM NRM NRM NRM SOA-bus Virtualization Orchestration Data center EMS Functional modules EMS VoIP Access CPE IPTV OLT Core VoIP Access IPTV CPE OLT Si Si SiSi Functional modules Metro (Ring) Metro (Ring) P2P Video (Mesh) Router P2P Video Si Si Si Si SiSi Si Si Si Si SiSi Sep. 21st, ‘08 22 Video-conference Video-conference
Packet ADM Packet ADM
Conclusion
Virtualization of network resources
Consideration of heterogeneity
O
API
Open API
SOA-based design for business process
Network resource management is a key
NRM and NRB
NRM and NRB
BPEL-based work flow management
Policy-based quality control
Scheduling
g
Functional modules virtualization
Lego ® block networking architecture
Customized networking
Future studies
Involvement of various functional modules
I t
ki
ith
t
k
t
Interworking with network resource management
This work is partially supported by NICT (National Institute of Information
and Communications Technology), Japan.
Sep. 21st, ‘08
23 2nd Workshop on FID