Department of Energy Office of Science ESCC & Internet2 Joint Techs Workshop Madison, Wisconsin.July 16-20, 2006 Network Virtualization & Hybridization Thomas Ndousse DOE, Office of Science Welcome

Department of Energy Office of Science IP Real-time Traffic burst Traffic Data Traffic voice Traffic wireless T1 Ethernet ATM SONET DWDM Satellite Video Streaming Soliton  The Internet – One of the greatest achievement of the 20-th century  The Internet must evolve, innovate, and scale to to meet increasing ubiquity  Ossification – Difficult to introduce new innovations in the network The Internet: The Evolving Global Network Infrastructure

Department of Energy Office of Science Original Design Expectations Current Operational Expectations  Research infrastructure  Commerce, defense, science,  Health, education, transportation, national security, etc  Best-effort, shared resources  Guaranteed bandwidth, dedicated channels, VLANs, VPNs, MPLS, GMPLS, quality of service  Twisted pair copper  Fiber optics, free-space optics, wireless, satellite, WiFi. WiMax  Single flow throughput: 9.6, 56kbps, 10 Mbps  100 Mbps, ~Gbps, ~Tbps  Transport stack: TCP  Zero-copy, off-load engines, RDMA,  Data traffic  Data, voice, streaming video,  Real-time, etc  Single flow throughput: 9.6, 56Kbps, 10 Mbps  100 mbps, ~Gbps, ~Tbps The Internet: then, Now, & tomorrow Simple System Complex System Ubiquitous System

Department of Energy Office of Science Network Virtualization  Virtualization attempts to solve the ossification problem by allowing multiple virtual networks to co-exist within a shared infrastructure Emerging Network Virtualization Why?  De-ossification - expand or eliminate the neck of the IP hourglass  Attempt to understand and manage large-scale network complexities  Stimulate research and innovations through deployment and testing of disruptive network technologies  Accommodate the requirements of diverse types of applications, especially high-end science applications

Department of Energy Office of Science PlanetLab – Larry Peterson, Scott Shenker, Jon Turner  A geographically distributed platform for deploying, evaluating, and accessing planetary-scale network services  A blueprint for future Internet  Platform for testing disruptive technologies before adoption GINI (Global Environment for Networking Innovations )  Beyond existing packet and circuit switching technologies  Building security into network architecture and protocol primitives  Towards a network science - Network architecture theories, the Internet as a complex system Overcoming Ossification and Complexity

Department of Energy Office of Science Topological Representation of Network Virtualization (PlanetLab) Source: PlanetLab and NSF GENI Workshop Report

Department of Energy Office of Science Layer 2 Vnet #1 Layer 4 Vnet #2 Layer 2 Vnet #3 Layer 3 Vnet #n Each virtual Net (Vnet) is an independent network that operates in a protected partition Inter-Vnet communication is allowed at the shared node Vnet #0 is designated as the network that performs the control and management Vnets can acquire and release resources to the main network infrastructure (dynamically?) Net #0 Control Net Expectations of Network Virtualization Main Network

Department of Energy Office of Science Layer 2Layer 3 Layer 2 Layer 1 Layer 5, 6, 7 Layer 1 Layer 2 Layer 3 Vnet #1 Vnet #2 Vnet #3Vnet #4 Control and Management Layer 4 Layered Approach to Network Virtualization

Department of Energy Office of Science Hybrid Network Model 1 1.Vnet is carved from the main backbone network 2.Resources can be exchanged between the Vnet and the main network Vnet (Circuit-Based) Main Network (IP-Based) Hybrid Networks Model 2 1.Vnet and the main network are two independent networks 2.No resource exchanged between the two networks Vnet (Circuit-Based) Main Network (IP-Based) Hybrid Network: Minimum Implementation of Network Virtualization

Department of Energy Office of Science A B F E D G C Routing Functions Switching Functions Hybrid Node Function InputOutput  Routing  Switching  Routing  Switching Hybrid Node Control InputOutput  MPLS  e-GMPLS  MPLS  E-GMPLS  e-GMPLS e-GMPLS:GMPLS with scheduling and reservation extensions Data Plane Technologies: MPLS, Layer 2 VLANs, Layer 3, SONET, WDWM Hybrid Network Node Functions and Capabilities

Department of Energy Office of Science Peering Issues of Hybrid Networks A A B F E D G C A B F E D G C Hybrid Net A Hybrid Control Plane  MPLS  e-GMPLS  MPLS  e-GMPLS Hybrid Net B Hybrid Control Plane  MPLS  e-GMPLS  MPLS  e-GMPLS

Department of Energy Office of Science Hybrid Network Research and Engineering Issues Hybrid Network Theory and Foundation  Hybrid network architecture  Hybrid network node capabilities, functions, signaling, and control  Hybrid network traffic engineering and network management Inter-Domain Issues for Hybrid Networks  Hybrid network taxonomy – Common terminologies and capabilities  Multi-domain control plane: engineering, policies, authentication  Hybrid network security issues: vulnerabilities of control planes, Circuit authentication  Hybrid network monitoring, performance measurement and prediction

Department of Energy Office of Science  Ossification is a problem across all federal agencies (DOE, DARPA, NSA, etc)  GENI is NSF’s view of the problem – A multi- agency approach to the problem is critical  Contribution from the science community is especially critical  LSN Workshop – The Future of Internet and Experiential Facilities Design, July 24-25, 2006 Inter-Agency View of Network Ossification & Virtualization

Department of Energy Office of Science Questions?