Comments
Description
Transcript
SDN and cloud testbed in Japan
JGN-X: SDN and cloud testbed in Japan Shinji Shimojo NICT/Osaka U 2012.11 New Generation Networking and its testbed SDN How does it help to improve cloud? Find some scinario on testbed in collaboraiton • • • • • • • • • • Future&Networks • • • $ • • • • • • • • 2011 • • Joint call with EU JGN-‐X Network infrastructure Overview New GeneraJon Network Layer （Partnership Service R&D Environment) DCN DCN DCN Plane DCN L2(VLAN)・L3(IP)NetworkLayer OF DCN OF Openflow PlaneOF RISE VN OF OF Physical L1(OpJcal Testbed) Network Layer VN Virtual Node Plane OF Sapporo VN VN VLAN Testbed Network Example StarBED3 1G 40G 10G 1G DF Sendai OF OF OF InternaJonal Circuit Hiroshima 10G Korea OF Fukuoka DCN 10Gx2 10G VN Kanazawa VN Okayama 10Gx2 VN VN 10G DCN Osaka 10G 40G NICT 10G Koganei VN 10G OF OpJcal Testbed DCN 40Gx2 40G Tokyo ■Koganei-‐Otemachi-‐Hakusan ■Oteamchi∼Otemachi OF InternaJonal Circuit VN Nagoya OF OF USA Wireless Testbed TH SG OF HK What is the StarBED (*BED) ? • It's an Internet/Ubiquitous System Emulator/Simulator • Re-configurable cluster supporting various user requirements • Large scale • over 1000 PCs 5 Shinsuke Miwa Concept of StarBED • Verification using Accurate Emulation • Verify Actual Running-Codes as in Operate, in Wall-Clock • PC, Embedded System, ...etc. • Large Emulation Capacity • • PC level: 1000 Physical nodes (over 10K using VM) Micro-sensor level: over 1million nodes • Configurable Network topology • • Flexible L2 topology configuration using Redundant network switches and VLAN L3 Routing emulations • PC-based routers • Installation of commercial routers at plug-in points • External connectivity External connectivities are available for both: • Control Path/Simulation Data Path • Connection to: • JGN-X (10GbE), JAIST (10GbE), and WIDE Project (10GbE) • 6 Shinsuke Miwa New Generation Network Technology • Network Virtualization • Programability on Network • ID/Locator Separation • Contents Oriented Network • In Network Processing • Optical Networking SDN= Wireless Networking • Software Defined Network Software Defined Network Control Plane Software Defined Network Control Plane Software Defined Network C control Control control Control Plane • Vnode/Corelab • Openflow • NSI path aggrigation for maintenance Control Plane • • • • SDN makes network design easy SDN gives you • • • programability of network design central control of all network device network virtualization Network Operation and configuration becomes centralized. Effective sharing of resources brings reduction of CAPEX&OPEX. Advanced network control such as • • • path aggregation for maintenance selective use of network appliance dynamic path addition for scale out Our challenge • Is SDN possible in wide area • Technology • Management • Business model RISE: OpenFlow/SDN Testbed • A large-‐scale OpenFlow/SDN testbed on JGN-‐X –for researchers, students, engineers, operators, … • Challenges –IniPally, demonstraPon of wide-‐area deployment of OpenFlow technology (since 2009) • Design of under-‐lay network architecture (tunneling) • Know-‐hows in shooPng OpenFlow network troubles –Now, development of OpenFlow testbed with … • User isolaPon that allows user-‐defined controllers • InterconnecPon of various testbed faciliPes such as network emulator, wireless/opPcal network testbed, cloud infrastructure, … 2012/7/17 Eiji Kawai 14 RISE infrastructure (v1.0) JGN-‐X links Q in Q trunk links OpenFlow-‐enabled vlan links mgmt and sec-‐chan seg. links E-OFS ksendai2 D-OFS kkanazawa2 ksendai1 kote2 nfukuoka2 ksendai-‐gs4k-‐1 kkanazawa1 kote1 nfukuoka1 kote-‐ex4200-‐1 kkanazawa-‐gs4k-‐1 nfukuoka-‐gs4k-‐1 kote-‐gs4k-‐1 nokayama-‐gs4k-‐1 note-‐gs4k-‐1 ndojima-‐gs4k-‐1 nokayama1 nokayama2 ndojima1 ndojima2 nnagoya-‐gs4k-‐1 nnagoya1 nnagoya2 note1 note2 Interop Tokyo 2012 DemonstraPon Malicious Traffic comes in Sampling Detector (1)Alert to (4)DROP Controller (2)COPY RISE OF Controller DPI (3)Detection by Deep Packet Inspection Demonstra,on Scenario: •Malicious traffic comes in, and sampling detector finds suspicious traffic flow •Detector sends alert to RISE OpenFlow Controller •RISE Controller copies suspicious flow to DPI machine •RISE Controller drops the flow 9 Virtual Cluster A VM VM VM VM VM VM VM VM Virtual Cluster B VM VM VM VM VM GRE AIST Open vSwitch VM VM VM VM VM VM Osaka U Open vSwitch Open vSwitch GRE GRE GRE Open vSwitch Open vSwitch UCSD VM GRE VM Open vSwitch Open vSwitch Openflow Controller Trema (Sliceable rouJng switch) GRE Openflow network Open vSwitch Open vSwitch Courtesy by Ichikawa VM VM VM DEMONSTRATIONS @ SC12 14 Nov. 2012 SC2012 18 Towards global OpenFlow testbed OFELIA RISE • Inter-‐connec,on among OpenFlow testbeds NDDI – RISE, OFELIA (EU), NDDI(US), and some Asian countries. • With OFELIA, starPng from using each other’s OF testbed by OpenVPN • With NDDI, RISE will inter-‐connect DemonstraPon at SC12 by uPlizing OESS Korea (Sapporo snow fes.) RISE Bangkok Singapore 14 Nov. 2012 SC2012 19 Integration of NSI and OpenFlow • ApplicaPon (Flow) based inter-‐domain route selecPon is demonstrated by Integrate OGF NSI, OpenFlow and PerfSONAR funcPons to realize mulP-‐domain SDN. • AutomaPc path selecPon based on PerfSONAR performance measurement. • Run three applicaPons (A-‐GOLE, SDN, Cloud IMS) at the same Pme. • Update Web services interface to NSI ConnecPon Service v2. • Introduce NSI Disocvery Service v1 to publish supported services. Future plan • NSI producPon service slate to start in JGN-‐X for provisioning inter-‐domain transport link. • ConnecPon and performance verificaPon of dynamic provisioned network. • Establishment of NSI and SDN integraPon technology. 14 Nov. 2012 SC2012 20 Cloud-‐based IMS 14 Nov. 2012 SC2012 KDDI 21 Collaborating with TWAREN Chicago RISE LA JGN-X Thailand JGN-X Singapore JGN-X KUAS NCU NCTU NTUST NCKU NCHC CHT-TL Planning to connect each SDN/OpenFlow testbed through US academic networks. (As of Nov.2012 : Connected via StarLight / PacificWave with L2 path) Smart X = Cyber Physical System • IoT (Internet of Things) • • • • billions of data sources, large scale data mobility, charn feature as a group or a set a new form of security and privacy Cloud Wearable device In house sensor medical sensor Mobile sensor heterogeneous sensor network Ongoing research project: Platform for the real-‐‑‒world analysis The aim of the project is to provide a plagorm for real-‐world analysis using various data sources such as sensors, contents of the social network services, etc. Smart LogisPcs MarkePng ObservaPons Environmental Monitoring New-‐generaPon ICT Services Crime PrevenPon Health cares Cloud Disaster Managements Multilayer Network by SDN Economic Accident CrimeOverlay Social Status Phenomenon Status Status Customizable Overlay Network Traffic Status Natural Disasters Analyze status / phenomenon in the realworld Combine various kind of information sources Analyze Contents Analyze Sensor Data Collect Contents Internet Collect Sensor Data 24 Global Testbed is the field Thank you Presented version is here! https://sites.google.com/site/sshimojo/talks