Grid vs. Cloud Computing and Why This Should Concern the Optical Networking Community



Similar documents
Cliquez pour modifier le style du titre

Dynamic Network Resources Allocation in Grids through a Grid Network Resource Broker

Flexible SDN Transport Networks With Optical Circuit Switching

ASON for Optical Networks

Using YANG for the Dissemination of the Traffic Engineering Database within Software Defined Elastic Optical Networks

SDN FOR IP/OPTICAL TRANSPORT NETWORKS

A New Fault Tolerant Routing Algorithm For GMPLS/MPLS Networks

Optimal. Data Center Connect (DCC) STRATEGIC WHITE PAPER

Juniper Networks NorthStar Controller

Communication Networks. MAP-TELE 2011/12 José Ruela

Alcatel-Lucent 1850 TSS Product Family. Seamlessly migrate from SDH/SONET to packet

Steve Worrall Systems Engineer.

Leveraging Advanced Load Sharing for Scaling Capacity to 100 Gbps and Beyond

100Gigabit and Beyond: Increasing Capacity in IP/MPLS Networks Today Rahul Vir Product Line Manager Foundry Networks

Relationship between SMP, ASON, GMPLS and SDN

MPLS-TP. Future Ready. Today. Introduction. Connection Oriented Transport

IP/Optical integration & Management. WDM Stuff AKA. Sami COPYRIGHT 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

ATRIUM: A testbed of terabit IP routers running MPLS over DWDM

SDN IN WAN NETWORK PROGRAMMABILITY THROUGH CENTRALIZED PATH COMPUTATION. 1 st September 2014

Research and Development of IP and Optical Networking

Virtualized Converged Data Centers & Cloud how these trends are effecting Optical Networks

Lightpath Planning and Monitoring

The Software Defined Hybrid Packet Optical Datacenter Network SDN AT LIGHT SPEED TM CALIENT Technologies

A Simulation Study of Effect of MPLS on Latency over a Wide Area Network (WAN)

MENTER Overview. Prepared by Mark Shayman UMIACS Contract Review Laboratory for Telecommunications Science May 31, 2001

Service Definition. Internet Service. Introduction. Product Overview. Service Specification

ICTTEN6172A Design and configure an IP- MPLS network with virtual private network tunnelling

An Emulation Study on PCE with Survivability: Protocol Extensions and Implementation

IP Core Transport Network

Recovery Modeling in MPLS Networks

Business Case for BTI Intelligent Cloud Connect for Content, Co-lo and Network Providers

Quality of Service for VoIP

NOBEL phase 2 - Large scale ASON/GMPLS network demonstrator

Workshop - New Paradigms for Routing in Optical Networks

An Architecture for Application-Based Network Operations

Please purchase PDF Split-Merge on to remove this watermark.

Using YANG for the Dissemination of the Traffic Engineering Database within Software Defined Elastic Optical Networks

Enterprise Network Simulation Using MPLS- BGP

Evolution of telecom network infrastructure for broadcast and interactive applications

Master Course Computer Networks IN2097

Broadband Networks Virgil Dobrota Technical University of Cluj-Napoca, Romania

Introducing Basic MPLS Concepts

MRV EMPOWERS THE OPTICAL EDGE.

DESIGN AND VERIFICATION OF LSR OF THE MPLS NETWORK USING VHDL

Transport for Enterprise VoIP Services

Project Report on Traffic Engineering and QoS with MPLS and its applications

MPLS is the enabling technology for the New Broadband (IP) Public Network

QoS Strategy in DiffServ aware MPLS environment

MPLS Layer 2 VPNs Functional and Performance Testing Sample Test Plans

IxNetwork TM MPLS-TP Emulation

THE SDN TRANSFORMATION A Framework for Sustainable Success

SDH and WDM A look at the physical layer

Carrier Ethernet Defined

EPIPE Connectivity Services

PART II. OPS-based metro area networks

How To Be Profitable With An Alcatel-Lucent 1655Amu System

Multiple Layer Traffic Engineering in NTT Network Service

Optical Software Defined Networking

Analysis of traffic engineering parameters while using multi-protocol label switching (MPLS) and traditional IP networks

ISTANBUL. 1.1 MPLS overview. Alcatel Certified Business Network Specialist Part 2

Connect & Go with WDM PON Ea 1100 WDM PON

MPLS-based Virtual Private Network (MPLS VPN) The VPN usually belongs to one company and has several sites interconnected across the common service

TRILL for Data Center Networks

Generalized MultiProtocol Label Switching

Virtual Private LAN Service

End-to-End Network Centric Performance Management

EVOLVING ENTERPRISE NETWORKS WITH SPB-M APPLICATION NOTE

An End-to-End QoS Architecture with the MPLS-Based Core

Enhancing BoD Services based on Virtual Network Topology Control

How To Provide Qos Based Routing In The Internet

GN3+ JRA1 Network Architectures for Horizon 2020

MPLS Environment. To allow more complex routing capabilities, MPLS permits attaching a

Enhancing Converged MPLS Data Networks with ATM, Frame Relay and Ethernet Interworking

Photonic Switching Applications in Data Centers & Cloud Computing Networks

Optical Networks for Next Generation Disaster Recovery Networking Solutions with WDM Systems Cloud Computing and Security

A Network Management Framework for Emerging Telecommunications Network.

SDN Testbed Experiences: Challenges and Next Steps

The Economics of Cisco s nlight Multilayer Control Plane Architecture

Business Case for Cisco SDN for the WAN

Carrier Ethernet Exchange. Robert J Huey Equinix, Inc. <rhuey@equinix.com>

Multi Protocol Label Switching (MPLS) is a core networking technology that

SDH and WDM: a look at the physical layer

David Tipper Graduate Telecommunications and Networking Program. Telcom 2110 Network Design, Slides 11. WAN Network Design

Performance Evaluation for VOIP over IP and MPLS

Boosting Capacity Utilization in MPLS Networks using Load-Sharing MPLS JAPAN Sanjay Khanna Foundry Networks

Policy-Based Fault Management for Integrating IP over Optical Networks

Connection-Oriented Ethernet: Operational and Deployment Considerations

Transport SDN Toolkit: Framework and APIs. John McDonough OIF Vice President NEC BTE 2015

MPLS Quality of Service What Is It? Carsten Rossenhövel EANTC (European Advanced Networking Test Center)

Transcription:

CAlcul Réparti sur Réseau Internet Optique à CApacité Surmultipliée* Grid vs. Cloud Computing and Why This Should Concern the Optical Networking Community OFC/NFOEC 2009 «Virtualizing and Scheduling Network resource for Emerging IT services: the CARRIOCAS approach» Pascale Vicat-Blanc Primet (INRIA) Dominique Verchère (Alcatel-Lucent Bell Labs) March 23 rd 2009 Optical Fiber Communication Conference and Exposition & the National Fiber Optic Engineers Conference * Distributed/Grid Computing services delivered by Internet Optical Networks with Ultra-high transmission Capacities Vx/N&N/E/09/xxx 1

Contents 1. CARRIOCAS Challenges and Project presentation 2. Efficient Network architecture combining capacity and flexibility 3. Scheduling, Reconfiguration & Virtualization (SRV) service 4. CARRIOCAS Pilot Network and Experimentations 5. Conclusion 2

CARRIOCAS Challenges To research, design and implement a ultra-high bit rate network reconfigurable according to grid applications connectivity requirements Transmission capacity 40Gbit/s per wavelength* Enhanced application-network interactions with SRV Guarantied network service deliveries in function of Application QoS requirements. To develop new Services & Usages collaborative engineering on virtual prototypes Computing Intensive Simulation Applications interactive visualization To specify new business models of services integrating IT access and connectivity The 40Gb/s transponder prototypes have been successfully deployed on Bruyères-le-ChâtelSaclay and SaclayClamart links, performance are at nominal level. 3

CARRIOCAS Partnership Optics Networking Computing Application Industrials SME s Academia Oct 2006-Sept 2009 LIST Financial partners: 4

CARRIOCAS Service optimized Optical Network dimensioning and sizing Network Traffic flow characterizations «Data Grid» generated by large organizations (companies and institutions) users (clients) and datacenters (IT resource providers) Relative Cost comparisons over 2007-2012 Tbps 6-40.00 5.5 5-35.00 4.5-30.00 4 3.5 3 Traffic Evolution - 25.00-20.00 2.5 2-15.00 1.5 1 0.5 ( ) (*) - 10.00-5.00 0-0.00 2007 2008 2009 2010 2011 2012 Router over Photonic X-Connect Year Carrier Ethernet Switch over Photonic X-Connect Router over SDH X-Connect Carrier Ethernet Switch over SDH X-Connect Router over WDM Carrier Ethernet Switch over WDM Carrier Grade Ethernet Switch over SDH X-Connect( ) TSS over ROADM: CARRIOCAS based architecture (*) Layer 2 (Carrier Grade Ethernet) and Layer 1 (WDM) reconfiguration capabilities Most Scalable Solution absorbing the increases of data grid traffic demands 5

Towards Virtualized Infrastructures and «Extended Service Providers» High Energy Physic engineers Bio-informatics Scientists Car Designers Manufacturers Usage SLA Cliquez pour Grid Service modifier le style du titre Service Storefront Service Integrator & Presentation Provider Connectivity Services Provider SLA Network Service Management (S.R.V.) Specific Software (AppsResources) Storage Resources CPU Resources Picture Walls (Visualization Scientific Resources) Instruments/Sensors (Acquisition Resources) CE Customers (Enterprises) UNI PE Provider Edge Node P P P Provider Node (P) (G)MPLS controlled Network Edge Nodes Telecom Infrastructure Operator P P Provider Edge Node (PE) PE UNI CE Customer Edge node (CE)!! IT infrastructures!! Datacenter Operators 6

CARRIOCAS services suited for Grid Applications V1 : Connectivity services Requirements: to guarantee Quality of Experiences for grid application users Adaptive connections configurations with QoS guarantee (bandwidth, latency, availability ) negotiated according to Grid Application workflow needs Technology enabler: GMPLS, PCE, Timed based Network management V2 : Full Virtualized services Requirements: To integrate grid resources management for executing the Grid Application workflows Selection, reservation and interconnection of Computational, Storage and Visualization Centers which may come from 3 rd party resource infrastructure providers (e.g. datacenters, HPC-centers) Co-selection and cross-optimization of network and grid (application) resources Technology enablers: WSRF, and network protocols + VXDL & VXcore (INRIA), 7

Binding of VISUPIPE flows (jobs) and Carrier Grade Networks services (resources) Visualization pipeline partitioning and mapping on Grid IT Resources Data Source Raw data Filtering Filtered data Data transformation: Volumetric transfer function, Surface construction Transformed data Rendering Frame buffer Display Cluster File System 24 Mpixels x 32 bits x 30 images/s x 2 for stereo = 46,08 Gb/s Low latency requirements for interactivity One Data flows can be mapped on one EVC allocated on dedicated WDM Automatic end-to-end Ethernet Virtual Circuit provisioning PSC TE Link WDM FA-LSP Dynamic mapping (CE-VLAN) Mapping (CE-VLAN) TSS Add/Drop Ports WDM LSP TSS Add/Drop Ports PSC [PSC,WDM] TE-link Transport Service Switch OCh [WDM,WDM] TE-link (Ingress ROADM) PSC OCh [WDM,WDM] TE-link [WDM, PSC] TE-link (Egress ROADM) Automatic Switched Optical Network Transport Service Switch 8

SRV position in Telecom Network Scheduling, Reconfiguration & Virtualisation Virtualized Services (v2) Service Service Negotiation Negotiation Service Service Publication Publication / / Notification Notification S.R.V. C in/out =20Gbps Service Provider s Carrier Network C Cliquez Patterns A pour modifier le style in/out =20Gbps Policies Patterns Business view Policies (WSLA) Contracts Contracts (WSLA) Service Service Selection Selection =40Gbps Composition B du titre Scheduling Scheduling Processes Processes SRV Processes Composition Shared Shared Information Information / / Data Data Model Model Resource Resource abstraction abstraction / / Virtualization Virtualization A B in/out Commercial Filters (policies) Connectivity Services (v1) Scheduled Management Components of Network Resources Network Management Function Layer Requests / Reponses Cds Network Management : Res. Scheduler Read / Write NMS Resource inventory A Scheduled Timed TED NetConf. PCE Consultation Planning tools E IGP-TE D Service Provider s Carrier Network C in/out A A in/out =20Gbps B in/out =40Gbps B Service Provider s Carrier Network C in/out A 3 A in/out =20Gbps 2 B in/out =4 0Gbps B Synchronization/ Connection Alimentation inventory TED =2 0Gbps C =20Gbps C Service Provider s Carrier Network A A in/out =20Gbps Service view Network view C in/out =2 0Gbps C Transport Networks B B in/out =4 0Gbps Element view B C 9

CARRIOCAS Demonstration platform CARRIOCAS pilot network demonstrates advanced network service functions and proving their concepts through an emulation platform Feasibility of automated network management and control with dynamic delivery of Telecom network services to end users: e.g. company customers. The first steps are focused on Connectivity services managed through a well defined Service Management System (SRV) described. Each connectivity service is designed with specific attributes including: Service Access points (UNI), throughput (x Gbps), end-to-end latency (y ms), time interval [T 1,T 2 ] GMPLS based control protocols, Timed based PCE, Network Resource Scheduler with Carrier Switches and Wavelength Cross-connects as connection services are based on Ethernet Virtual Circuit over Wavelength connections: Provisioned: through NMS with automated network Operations through MTOSI 2.0 (triggered by SRV) Signaled: RSVP-TE provides dynamic E2E guaranteed QoS on two routing areas 10

Service Activation Interface : SRV-NRS Connection Service Component Activation: EVC Services SRV NRS (phase 1): Timed EVPLI, EVPLAN, E-TREE provisioning TE management Connection Provisioned Monitoring NRS SRV-DB Background and periodic process (from Network States) Downward information: SRV NRS Connectivity topology is an Ethernet Service graph Access points: location, Bandwidth (pipes) Latency on the links Commands for new connectivity services Feasibility Reservation Provision Activation Upward information: NRS SRV Connection status: Reserved / Available Provisioned / Unprovisioned Activated (i.e. used) / De-activated Shared Info/Data Model based on MTOSI 2.0 / TMF SCAI: Resource facing Interfaces (Connection data model) 11

References on Experimentation plans Ethernet MPLS-TP over WDM Network architecture referencing : Reconfigurable-OADM based WDM transmission: 1626LM (with 40 Gbps) Carrier Grade Ethernet based Transport Networks: 1850TSS Network Configurations based on Management: 1350-OMS service functions Large Scale Distributed Applications (Grid Applications): Computing and Storage Service on Demand (Ter@tec): LUSTRE CFS Application Instruments: Industrial (Remote Visualization): EDF Picture Walls North Bound Interface Negotiation protocol (e.g. WS-agreement based) Connection Services Requests Connectivity Services Scheduling SchedulingReconfiguration Reconfiguration &&Virtualization Virtualization (SRV) (SRV)service service Delivery Network Management Systems (1353NM 1354BM-Ethernet) Advertisement protocol (e.g. WS-notification based) South Bound Interface Application Client Visual. Server Network Controller Ethernet Switch (Customer Equipment) Network Controller Grid Application Client / Server Ethernet Switch (Customer Equipment) 1850 TSS-320 (Provider Edge) b/s nx10 G : λ on link 2 smissi n a r t y t i c a p a High-C EDF Clamart (PoP 1) A1850 TSS-320 (PE) 1626LM CE 1626LM 1 PE CEA-Saclay (PoP 3) 1626LM link 4 Application Client Storage Server PE link 3 s: ission Transm Gb/s 40 s / b G 0 λ: 4x1 CE 10Gb/s Interfaces (Ethernet frame aggregation over λ) HPC + Storage Servers Network Controller link Network Controller Univ. Orsay (PoP 2) CEA Ter@tec (PoP 4) 1626LM 12

CARRIOCAS Network Pilot and test bed configuration 4 x 10GEth. 1354BM Network Management 1353NM Picture wall Clamart Local monitoring Issy les Moulineaux Cliquez pour modifier le style duswitch titre Transport Service Carrier Grade Ethernet with T-MPLS capabilities 1850 TSS Trib 40G prototypes fournis gracieusement par Alcatel dans SP1 durant la vie du projet 4 x 10GEth. 4 x 10GEth. Local monitoring storage Saclay Orsay 4x10 GEth. Access storage Légende 40 Gbps Transmission 1626 LM Super-computers Ter@tec Bruyères Le Châtel Lambda 1 à 10G omnibus Lambda 2 à 10G Lambda 3 à 10G Lambda 4 à 10G Lambda 5 à 40G Nx10 Gbps Transmission 4 x 10GEth. Deployment phase 1 (10Gb/s): June 2007 Deployment phase 2 (40Gb/s): June 2008 13

Towards Network Centric Cloud Computing Respond to strong demands by Organizations on accessing High- Performance IT services remotely through WAN with strict QoS: Grid Cloud Computing towards Utility Computing Service optimized Transport Networks & Data Grid Facilities will be built on different requirements to offer e-science services Depending if Public or Private owned infrastructures Needs to coordinate our efforts for standardizations SRV v1 software, developed by INRIA, automates: Support for different network services offered by multilayer networks (Lx-VPN) Mapping of applications/users requirements over connectivity services Fine grain Dynamic Bandwith Requests scheduling Connection service Reconfiguration Optimization X-optimization Standardization efforts at ETSI (Grid TC) led by BT and Alcatel- Lucent with contributions from BT, FT, Telefónica, Nokia-Siemens, close follow-up at Open Grid Forum (through Phosphorus collaboration) 14

Many Thanks to CARRIOCAS-SP2 Colleagues for their contributions French Ministry of Industry, Essonnes, Haut-de-Seines and Paris General Council for their supports on CARRIOCAS project http://www.carriocas.org Special thanks to PHOSPHORUS and G-Lambda projects for their stimulating and valuable interactions Questions? 15

CARRIOCAS Network Environment Emulated Platform is based on CARRIOCAS network pilot architecture* Two Routing Areas (1 per switching layer): (i) CGEthernet (ii) Wavelength 5 PoP emulated, each PoP is Carrier Switch (Ethernet-switching capable) + PXC (Lambda-switching capable) 10 Node controllers (5+5): RSVP-TE + OSPF-TE 2 Path Computing Engines (1 per layer): PCEP, OSPF-TE 1 NMS: performs resource inventory at L2 and L1 Dynamic Ethernet LSP establishments 5 EVPLI services 1 EVPLAN service (4 locations) Static Wavelength LSP characterized by Capacity: 3x10Gbps, 1x30Gbps, 1x40Gbps Latency: 1ms t 10 ms * Mais contient des fonctions supplémentaires Ingress node LSR1 λ / GMPLS Routing Area 0 OADM node Carrier Grade Ethernet Network services Routing Area 1 Virtual Topology Wavelength X-Connect Ethernet LSP 1 Ethernet LSP 2 Wavelength connection 1 CORE Intermediate edge node Egress node LSR2 Wavelength connection 2 UNI 16

5 point-to-point Ethernet LSP (E_LSP): E_LSP(ERO,BW(Gbps),delay(ms)) E_LSP_1 ({AB},5,1); E_LSP_2 ({ABC},10,5); E_LSP_3 ({A C},40,1); E_LSP_4 ({AC D},20,10); E_LSP_5 ({A B E},15,5); 1 point-to-multipoint Ethernet LSP: SRV-DB E_LSP_ 6({tree_1},10,10); SRV Service Scheduling MTOSI Commands (over HTTP/SOAP - AXIS) 6Network configuration implementation (1/2) B D LSR_A RSVP-TE + OSPF-TE LSRs Layer 2 Switch Capable (51) TE-link LSR_D A E C LSR_C tree_1 NRS Agent Network Management System XML NetConf RPCs PCC PCEP Commands (RPCs) S-PCE PMSER algorithm S-PCE TE-Link DB TE-Link DB LSR_E LSR_4 LSR_B LSR_3 LSR_1 RSVP-TE + OSPF-TE LSRs Lambda Switch Capable (150) TE-link LSR_2 LSR_5 17

Network configuration implementation (2/2) 5 wavelength LSPs (L_LSP): L_LSP(ERO, BW(Gbps) L_LSP_1 ({1 4},10); L_LSP_2 ({1 (4) 3},30); L_LSP_3 ({1 (5) 2},40); L_LSP_4 ({2 3},10); L_LSP_5 ({1 5},15); Scenarios Connectivity services managed by SRV T 0 =1 (January 1 st ), T 1 =55 (February 24 th ), duration=nb of slots (1 time_slot = ½ day), e.g. duration = 60 (30 days) End-points T 1 /duration Max. Throughput/ Guaranteed BW. e2e delay (ms) Service availability Application_id (information) A B 23/10 38.4 / 30 Gbps 5 ms 99.999% Ultra-high perf. Visu Coll. AB 26/60 10 / 5 100ms 99.999% Back-up Storage A D 24/20 20 / 10 5 ms 99.999% Ultra-high perf. Visu E D 38 / 6 5 / 1 1 ms 99.999% Video- Streaming D C 35 / 12 30 / 10 10 ms 99.9 % Data Storage C{A,E,D} 30 / 16 15 / 10 5 ms 99.99% Tele-Call-Conf T 0 T 1 -ε T 2 +ε T 2 = T 1 + duration 18