ITU Kaleidoscope 2013 Building Sustainable Communities Design and implementation of virtualized ICT resource management system for carrier network services toward cloud computing era Yoshihiro Nakajima, Hitoshi Masutani, Wenyu Shen, Hiroyuki Tanaka, Osamu Kamatani, Katsuhiro Shimano, Masaki Fukui, and Ryutaro Kawamura NTT Network Innovation Laboratories nakajima.yoshihiro@lab.ntt.co.jp Kyoto, Japan 22-24 April 2013
Contents 1. Background & Motivation 2. Contribution 3. Requirements for virtualized ICT resource management 4. Carrier network service architecture design 5. Implementation for mobile network service 6. Conclusion
Background and motivation High flexibility in provisioning, control, and management is indispensable to satisfy new type demands of customers in cloud computing era e.g. Speed to market, service elasticity for traffic demands, cooperation of NW and IT resources Network carriers face to reduce OPEX/CAPEX while ensuring variety, reliability, availability, and management flexibility with shared ICT resource Disaster-tolerant network services Advanced technology in ICT resource virtualization Virtualization middleware, cloud computing in IT Software-defined networking/openflow in NW
Approach Management Engine: Redesigned software component-based management system + Virtualized NW/IT resources for carrier network service Virtualized ICT resource information modeling for cooperation of NW and IT resources Decoupling the functionality from the entity of the physical ICT resources Software defined networking Matured IT virtualization technologies
Contributions A virtualized ICT resource information model is designed to take advantage of ICT virtualization technology in carrier network service Designing a modular/layered management system architecture for virtualized ICT resource Flexible service cooperation to ensure dynamic resource accommodation between services with virtual ICT resources is demonstrated Future standardization and improvement issues in the virtual ICT resource management field are summarized
Existing network service architecture: NGN and Internet Application service functions and network functions are separately managed Cooperation functions between application service and ICT resource are limited Poor support for global optimization NGN ASF SCF Internet AS TCF UE TF TF Other Internet UE TF UE: User Equipment ASF: Application Support Functions SCF: Service Control Functions TCF: Transport Control Functions TF: Transport Functions TF Other NGN UE: User Equipment AS: Application Servers TF: Transport Functions
Approach portfolio of virtualized resource management Resource Resource Area Type Network Resource (e.g. Carrier Network) IT Resource (e.g. Data Center) Physical resource Conventional Approach for Network Service, Architecture, Management, and Modeling Conventional Approach for Data Center Service, Architecture, Management, and Modeling Network Virtualization Technologies Server Virtualization Technologies Virtual resource New Approach for Virtualized Network Service, Architecture, Management, and Modeling Cloud Computing Approach for Cloud Computing Service, Architecture, Management, and Modeling New Approach for Virtualized Network and IT Service, Architecture, Management, and Modeling This study s Focus area/issues
ITU-T Y.3011: NW virtualization for future NW Various services run over logical isolated network partition (LINP) using shared virtualized network resource Service2 Service4 Service1 Various Services LINP1 Virtual Networks LINP2 Service3 LINP3 LINP4 LINP1 Manager LINP2 Manager LINP3 Manager LINP4 Manager Four main issues Coexistence of multiple NW Simplified access to resources Virtual Resources Physical NW1 Physical NW2 Virtual Resource Manager Physical NW1 Manager Physical NW2 Manager Flexibility in provisioning Evolvability Physical Resources Physical NW3 Physical NW3 Manager
Requirements of virtualized ICT management for carrier network Virtualized ICT resource information model Extended/integrated resource information model for virtualized NW and IT resource cf cloud technologies for virtualized IT resources, ITU-T Y.3011 for virtualized NW resources Additional requirements to ITU-T Y.3011 Multi-tenant support Cooperation control and resource provisioning for virtualized ICT resource Traceability and mapping relationship information between virtualized and physical resources
Concept design of the proposed network architecture Encourages flexible network service operations and dynamic service reconfigurability for service providers and carrier operators Public northbound API New neck of hourglass (cf IP)
Management system design Layering/modularizing architecture For long-term evolution and rapid development Utilizing software component technologies Leveraging software-defined networking technologies Decoupling systems functionality from physical resource entities with programmability and ICT resource virtualization Using function-aware ICT resource abstraction model and model-based control
Architecture overview of Management Engine Converts request into resource allocation or resource control request Handle differences between model and actual ICT equipment Search optimal combinations and multilayer-aware resource coordination/control Management policies ICT resource coordina on Management logic Management engine Resource coordina on broker Network resource manager Device controllers Resource compiler Abstracted NW/IT resource model Changes in network or IT resources IT resource manager Network resources (e.g., Call control node, OpenFlow switch) Open innova on with so ware components OSGi bundles Bundle repositories U liza on of bundles IT resources Mobile terminals
Function-aware virtualized ICT resource information model Unified information model based on ITU-T G.805, TM Forum SID, and NDL for ICT resources Allows multi-layer-capable & ICT coordination processing Consists of function block, layer adaptation, and service U_IF1 Switching Table Resource parameter D_IF1 D_IF2 ingress port ingress label egress port egress label U_IF1 X D_IF2 X D_IF1 Y D_IF2 Z paramete r switching capacity value N (G) function block NW:IP,MPLS,VLAN IT: storage, computer inter-layer connection Service Masking fine details Layer adaptation function block (upper layer) function block (lower layer) upper label X Z function block (lower layer) lower label ITU Kaleidoscope 2013 Building Node Sustainable Communities M Node
Architecture implementation for mobile network services Voice and Video service run on prototype network designed by referencing architecture of 3GPP LTE IMS with dynamic scale-out/scale-down extension for voice Flash over HTTP for video Management Engine dynamically controls both OpenFlow network and virtualized IT resources
Configuration and its information modeling Physical configuration Two DC, Three APs:P- CSCF of IMS and HTTP video server on three VM One backend DC: {S,I}-CSCF of IMS OpenFlow network Information modeling Layer-4-level path as minimum elements of NW virtualization view Underlying resources are systematically configured 202.12 204.12 201.11 203.11 205.12 206.12 AP1.1 192.168.1.0/24 AP2.1 192.168.2.0/24 AP3.1 192.168.3.0/24 Clients 192.168.0.1/24.2.1 Node#1 VoIP Client 2 Flow VoIP Client 1 OFS #2 OFS #1 OFS #3 OpenFlow Network IMS (Proxy-CSCF) 192.168.100.0/24 OpenFlow SW OFS #4 OFS #5 Service connec vity OpenFlow #1 Flow OFS #6 Data Center -A VM#1.111 VM#2.111 VM#3.111 Data Center -B.121 OFS #7 VM#1.111.122 VM#2.111 VM#3.111 Video on demand server OpenFlow SW User-defined Label Switching (OF) OpenFlow #2 Flow.111.112.113.123 Backend Data Center IMS (S-CSCF, I-CSCF).1 192.168.110.0/24 Data Center -A Server 192.168.100.1 DC-A Node#4 SIP (P-CSCF) Flow Physical NW ETH ETH ETH ETH ETH ETH
Management Engine Implementation All modules are developed as OSGi bundles ICT-tightly-coupled resource coordination and function assignments are performed with multilayer resource coordination search engine OpenFlow driver controls layer-3/4 flow as label switching on OpenFlow switch Voice Video Service operator s view Client Operator view Topology view OSGi Bundle Viewer REST REST REST Management Logic (Tomcat) Server Topology view service layer virtual network layer REST Management Engine (Equinox) Resource compiler for mobile Multilayer Resource Coordination Search Engine (TSE) Flow-based NW topology DB OF Driver(OpenFlowj) VM/process control driver SQL Abtracted ICT resource model( PostgreSQL) OpenFlow Kyoto, Japan, 22-24 physical April layer 2013 OpenFlowSwitch OpenFlowSwitch KVM Custom command
Operation strategy policy Strategy 1: Simple management Maximize the performance of video-ondemand service Resource allocation for QoS guaranteed voice service NW link failures and burst demand for voice service Switch policy strategy to Strategy 2 Strategy 2: Intelligent management Preferential NW and IT resource allocation for voice service Scale-down of Video on-demand service Voice service Video-on-demand service
Demonstration scenario: disaster scenario 1. Normal state: Strategy #1 2. Network problems due to a major earthquake link failure occurs and burst emergency call demands for voice service 3. Enforcement for voice service with network restorations Strategy changeover from #1 to #2.
Demonstration link to be disconnected Voice service (R-CSCF) Video-on-demand service Strategy changeover from #1 to #2 Configuration is changed in one minute path restoration & enhancement for voice service scale-out of voice service scale-down of video service disconnected link
Future research area and standardization issues Unified abstracted ICT resource information model Realizes carrier-grade network control and management from physical entity to the service level for Standardized API for virtualized ICT resource control and provisioning Lowers entry barrier for new service creation Refers API of cloud middleware and DMTF Operational policy description language enables arbitration of resource allocation among service and automatic service management
Conclusion Management Engine, a virtualized ICT resource management system for carrier network services is designed and implemented Function-aware virtualized ICT information model Demonstration shows its flexibility and capability Future work Standardizing a virtualized ICT resource information model