ARCHSTONE Advanced Resource Computation for Hybrid Service and TOpology NEtworks New Projects Kick-Off Meeting Fermilab, Batavia, IL September 28 th, 2009
Introduction Network architectures and services - Increasing Complexity • Growing need to consider many more dimensions (or constraints) for control and provision of network resources • User services becoming more complex and varied - virtual organization specific resource control, on-demand creation of network topologies across multiple layers, extended user-to-network "conversations" Essentials for multi-layer network control • Next-generation networks tend to be architected as a heterogeneous “multi-layer, multi-technology” construct • Multiple types of services simultaneously over common infrastructures • Need to hide network details from users, i. e. , “services virtualized” Where are we now? • Initial renditions of user requested service provision paradigms have been realized in DOE ESNet SDN, Internet 2 DCN, and others. • Challenging issues unresolved: rich service interface definition, dynamic topology computation, multi-layer control etc.
Personnel • • • USC/ISI • Tom Lehman • Xi Yang ESnet • Chin Guok • Eric Pouyoul • Inder Monga UNM • Nasir Ghani • Kevin Xie
Multi-X Network Control and Service Provision Heterogeneous network environments • Networks are multi-service, multi-level, multi-layer, multi-technology, multi-vendor, multi-domain, multi-provider and multi-policy • Multi-X offers various networking options in terms of physical interface types, transport technologies, performance characteristics etc. • It also means expanded computation space and complex constraints. Key approaches to multi-X network control and provision • Multi-layer resource “slicing” by vertical integration - Network resource computation with inherent multi-X capability - Power and intelligence to generate multi-layer virtual network topologies, circuits, and/or partitions on-the-fly • Service-oriented networking - Advanced network service interface framework for higher-level services and applications - General purpose multi-layer topology computation element, independent of computing environments, applications and operation models
Network Service Interface (NSI) The NSI framework defines multi-layer network services that hide network operation details and allow for on-demand “virtualization” or “slicing. ” NSI Parameters by Categories Path TE Parameters (source, destination IDs, bandwidth, link switching type, etc) Layer-Specific Params (MTU, VLAN Tags, SONET VC type, WDM wavelengths etc. ) Qo. S Parameters (bandwidth guarantees, latency, jitter, packet loss etc. ) Scheduling Params (start and end times, ranges of acceptable time windows, priorities) Routing Profile (explicit routes, link inclusion and exclusion lists etc. ) Execution Method (‘hard’, ‘soft reservation’, (‘hard’, ‘soft ‘query only’, ‘schedule only’, ‘preemptible‘ etc) Protection Capability (‘non-protection’, ‘ 1: 1’, ‘ 1+1’, ‘shared mesh protection’ etc. ) Management Info. (auxiliary information for service monitoring and trouble shooting) AAA Information (parameters for user authentication and policy enforcement) Value-Added Services Basic Operations request for single path Batch Operations request for a group of paths as atomic operations Conditional Operations operations with conditional logic to support workflows Topology Services request for custom topology views and information A “service” is defined by elements containing some of the above parameters. NSI services can be flexibly integrated into virtualization workflows and be discovered and invoked via standard mechanisms, such as “Web Services. ”
Multi-Layer Networking - Data Planes Many Data Plane options to provide Hybrid Network Services: • Layer 3 for PSC with Qo. S (IP Routing, MPLS) • Layer 2. 5 for PSC with Qo. S (MPLS, e. g. Ethernet over IP) • Layer 2 for L 2 SC (often Ethernet) • Layer 1. 5 for TDM (often SONET/SDH) • Layer 1 for LSC (often WDM switch elements) From client hand-off perspective the "service" of choice is currently an Ethernet service • This is independent network technology layer • That is, all of the above Data Plane technologies can encapsulate Ethernet for the client service interface Similar approaches will be seen in the future for multiple client "services" • Infini. Band • OTN/G. 709 • Fibre Channel • Others
Multi-Layer Networking - Architectures • Many ways to architect the Data Plane Parallel Combined Multi-Layer Multi-Service Multi-Layer Hybrid Service
Service-Oriented MX-TCE Architecture The core capability of the project is Multi-X Topology Computation Element (MX -TCE) • Handling multi-dimensional information and constraints that are typically not considered in current network PCE engines. • Computing both paths and topologies for multi-layer networks • “NSI ready” to enable general-purpose network resource computation. The service-oriented design allows MX-TCE to be flexibly incorporated into various virtualization systems.
MX-TCE Capabilities Ultimately must be Inter. Domain Boundary Topologies will need to reflect options for multi-domain provisioning User Specific Multi-Domain Topology Layer 3 Layer 2 Topology Domain Boundary Layer 1 Layer 3 Topology Layer 2 Layer 1 Path Computation Time Domain AAA Management (SNMP) Data Administrator Requirements PCE to PCE Coordination Time Domain AAA Management (SNMP) Data Administrator Requirements
Project Goals Dynamic management and control of multi-X networks • Develop capabilities to seamlessly manage and control network resources across all network layers • Develop service-oriented path and topology computation capabilities for general-purpose network virtualization applications Test and develop capabilities in lab • Develop technologies with focus on emerging R&E networks • Use DOE Testbed for development and testing • Seek for acceptance of the NSI framework by R&E networking organizations and broader communities Test and deploy capabilities into production networks • Integrate MX-TCE with existing network provisioning systems, particularly OSCARS • Deploy MX-TCE in other production environments Enable other uses and deployments for experimental and research projects
Partner Institutions and Roles USC/ISI • Lead institution for management and execution of the project, responsible for overall architecture, design and implementation issues for MX-TCE • System requirements, capabilities, use cases and architecture definition • Software development and implementation ESNet • Primary driver of the research activities, responsible for identification of system requirements, capabilities, use cases and architecture definition • ESnet Production operations integration • Software development and implementation UNM • Theoretical research on path computation algorithms and techniques. • Lab performance test and transform theoretical work into implemented components
Project Outline Advanced network service interface • • Requirements definition, architecture development, design and implementation Protocols and formats for service requests and answering complex questions regarding multi-X capabilities, states and service provisioning status. MX-TCE design and implementation • Develop ability to i) operate in multi-layer, multi-technology, multi-vendor environments, ii) integrate multi-dimensional information from all planes, iii) incorporate information based on specific network architectures and requirements, and iv) utilize advanced algorithms within complex computation space. Develop interfaces and workflows to allow/assist other advanced capabilities to utilized features enable by MX-TCE and Advanced NSI • network virtualization, topology instantiation, protection/restoration, IP routed traffic engineering/traffic grooming, control plane and network management integration, AAA/security architecture Technology transfer • Ensure that there is a clear evolution (integration) path from current multi-level/multi-domain network control frameworks to the advanced architectures and features to be developed under this project.
MX-TCE Integration with OSCARS
MX-TCE in Broader World MX-TCE provides flexible and efficient options for any “NSI-ready” service provisioning systems in multi-layer / multitechnology networks Ask MX-TCE acts as network resource broker for virtual topology slicing and instantiation in any “NSIready” computing virtualization and grid computing systems Give me an IP overlay triangle topology with 500 Mbps per link. Can I have a slice of 1 Gig. E VLAN with customer tag 300? I prefer paths with least jitter. What are available 8: 00 -10: 00 AM every Friday? Site A Site B multi-layer network Site C Application/User/Virtual Organization Unique Topology
Project Annual Schedule Year 1 Year 2 NSI requirements document (Q 1 -Q 3) NSI architecture and design document (Q 2 -Q 4) Multi-dimension TEDB architecture and design (Q 2 -Q 4) Initial AAA information model and interface design (Q 2 -Q 4) Initial NSI standards body coordination (Q 3 -Q 4) Initial MC-TCE architecture and design (Q 3) Initial path and topology computation architecture and design (Q 3) Initial multi-dimension TEDB implementation (Q 4) Initial topology filter and graph transformation/construction study (Q 4) Initial path and topology algorithms study (Q 4) Completion of NSI documents and standard body coordination (Q 1 -Q 3) NSI client reference implementation (Q 1 -Q 3) Multi-dimension TEDB architecture and design, and implementation (Q 1 -Q 4) Path and topology architecture and design, and implementation (Q 1 -Q 4) Multi-layer, multi-technology provisioning architecture and design, and implementation (Q 1 -Q 4) Multi-layer virtualization topology architecture and design, and implementation (Q 1 -Q 4) AAA information model and interface design and implementation (Q 1 -Q 4) Initial architecture and design for protection and restoration mechanism, IP routed traffic engineering and grooming, and control plane and management plane integration (Q 1 -Q 4) Technology transfer plan (Q 2 -Q 4) Year 3 Multi-layer network virtualization topology instantiation implementation (Q 1 -Q 4) Protection and restoration mechanism architecture and design, and implementation (Q 1 -Q 4) IP routed traffic engineering and grooming architecture and design, and implementation (Q 1 -Q 4) Control plane and management plane integration architecture and design, and implementation (Q 1 -Q 4) Technology transfer plan and deployment (Q 1 -Q 4)
Project Deliverables Advanced network service interface • NSI requirements document and architecture and design document • NSI description and documentation submission to standards bodies, including Open Grid Forum (OGF) and Global Lambda Interchange Facility (GLIF) • Sample NSI reference client implementation MX-TCE design and implementation • MX-TCE architecture and design document and implementation - Topology data incorporation in multi-dimensional TEDB - Path and topology computation - Multilayer/multi-technology provisioning - Service interface with production systems such as OSCARS • The implementation will develop specific methods/techniques to incorporate data from ESNet Juniper routers and Infinera WDM switches and enable multi-layer/multi-technology provisioning in production networks such as ESNet. • MX-TCE test and report that combines the above to generate a complete MX-TCE prototype.
Project Deliverables Interface documentation, reference implementations, test reports for how other advanced capabilities can utilize features enable by MX-TCE and Advanced NSI • Advanced multi-layer network virtualization and topology instantiation • Advanced multi-layer IP routed traffic engineering and traffic grooming • Advanced multi-layer protection and restoration mechanisms • Control plane and network management integration • Multi-X network AAA and security architecture Technology transfer • Technology transfer plan document that outlines the transferring process • Deployments to build and test the developed technologies on production networks
Interaction with Related Projects Interaction with the other technology development projects is very important to ensure the new class of "network infrastructure" services which are developed are correct in terms of: • features • interfaces • degree of exposure/abstraction of physical devices • ability to incorporate into useful workflows
Thank You! Discussion
Project Outline Develop interfaces and workflows to allow other advanced capabilities to utilized features enable by MX-TCE and Advanced NSI: • Advanced multi-layer network virtualization and topology instantiation - Develop ability to request and instantiate whole network topologies based on the MX-TCE capabilities. - Develop “Vendor Interface Modules” for specific testbed and production network operations, including 100 G Testbed, ESNet Lab, ESNet Production and others. • Advanced multi-layer protection and restoration mechanism development - Develop ability for advanced multi-layer protection and restoration schemes to utilize the MX-TCE capabilities. • Advanced multi-layer IP routed traffic engineering and traffic grooming - Develop ability for advanced multi-layer IP traffic engineering and traffic grooming schemes to utilize the MX-TCE capabilities.
Project Outline Develop interfaces and workflows to allow other advanced capabilities to utilized features enable by MX-TCE and Advanced NSI (continued): • Control plane and network management integration - Develop the technologies and mechanisms to allow for control plane and management plane coordination for monitoring and debugging multi-layer network services, capabilities and states based on the MX-TCE capabilities. • Multi-X network AAA and security architecture - Develop the architecture, information models and technologies to incorporate AAA and security frame into the MX-TCE design and associated multi-layer network services. - Develop a framework to flexibly absorb AAA information from various sources.
Скачать презентацию ARCHSTONE Advanced Resource Computation for Hybrid Service and
60e9ba40cb54a9dc435d835aa4117f84.ppt