FUTURE INTERNET Testbed @TWAREN Che-Nan Yang NCHC,Taiwan Overview - PowerPoint PPT Presentation

FUTURE INTERNET Testbed @TWAREN  Che-Nan Yang NCHC,Taiwan

Overview  OpenFlow Testbed in TWAREN  HPDMnet Multicast Streaming with OpenFlow  Future Work 2

Future Internet  There are many serious limitations in current Internet.  Scalability  Security  QoS  Virtualization  Future Internet is a summarizing term for worldwide research activities dedicated to the further development of the original Internet. (From Wiki) 3

Future Internet Testbed  For innovations and researches in Future Internet, the testbed requires some advanced concepts:  Programmability  Virtualization  End-to-end slice 4

OpenFlow  Make deployed networks programmable  Makes innovation easier  No more special purpose test-beds  Validate your experiments on production network at full line speed 5

OpenFlow Testbed in TWAREN NOX iCAIR Capsulator OpenFlow Switch TWAREN OpenFlow Capsulator OpenFlow Network Capsulator Network @KUAS @NCKU We do not have pure Layer2 network in TWAREN, using the Ethernet- in-IP tunnel instead. 6

Slice Isolation Problem (1/2) T r a f f i c b e t w e e n N C K U & N C H C S A U K o t s r o r r i c m i f f o a s r l t a s c e C i f e f H a s C r T o N s l a & S U A K U C K N n e e w t e b 7

Slice Isolation Problem (2/2) Capsulator @ NCHC Capsulator @ NCHC Tunnel port Border port Border port Tunnel port Border port Border port User space thread thread1 thread2 thread thread1 thread2 Kernel space Listen all the packets Listen the MAC-in-IP from border port packets Both threads listen the common physical interface eth0 eth1 eth0 eth1 Tunnel port Border port 8

Capsulator + Open vSwitch (1/2) Capsulator @ NCHC Capsulator @ NCHC Border port Border port Border port Border port Tunnel port Tunnel port User space thread1 thread2 thread1 thread2 thread thread Kernel space tap tap tap tap 0 1 0 1 Flow Table Open vSwitch Open vSwitch eth0 eth1 eth0 eth1 kernel module kernel module Tunnel port Virtual border port border port 9

Capsulator + Open vSwitch(2/2) OvS-controller OvS-controller OpenFlow protocol OvS-openflow OvS-openflow Capsulator Capsulator daemon daemon Using the OpenFlow- Using the OpenFlow- tap0 tap1 enabled tap0 tap1 enabled Open vSwitch to isolate Open vSwitch to isolate flow-based switching traffic in both slices traffic in both slices Open vSwitch Open vSwitch kernel module kernel module eth0, eth1: physical interfaces tap0, tap1: virtual interface eth0 eth1 eth0 eth1 system call interface bridge interface 10

Ethernet-in-IP Tunnel 11

OpenFlow Testbed with TWAREN VPLS (Scheduled) OpenFlow Network @NCHC OpenFlow TWAREN Switch VPLS OpenFlow Network @KUAS OpenFlow Network OpenFlow @NCKU Switch OpenFlow Switch 12

TWAREN International Circuit 13

International OpenFlow Testbed 14

International GENI (iGENI) Testbed

booth Video Streaming over High Performance Future Internet HPDMnet HPDMnet s b p G 1 Streaming Client 10Gbps s p b M 2 2 iCAIR booth TWAREN 6 TWAREN OpenFlow OpenFlow booth Testbed Testbed 1 G b 1Gbps p s booth 1Gbps booth Streaming Server NCKU KUAS Streaming Server Streaming Server 16

HPDMnet Overview  An International Consortium of Research Centers Has Formed a Cooperative Partnership To Address Key Challenges and Opportunities Related to Using Dynamically  Provisioned Lightpaths for High Performance Digital Media (HPDM)  Multiple Sites Require High Performance/High Volume/High Definition Digital Media Streaming Simultaneously Among All Locations (Point-To- Multipoint, Multipoint-To-Point, Multi-Point to Multi-Point)  This Consortium Is Designing and Developing New L1/L2 Capabilities That Can Provide Large Scale HPDM Services, Which Can be Used for Any Data Intensive Application, Not Just Digital Media 17

HPDMnet Consortium Member CANARIE  Communications Research Centre (CRC) Canada  Electronic Visualization Laboratory(EVL), University of Illinois at Chicago  I2Cat  Inocybe  Institute of Computer and Network Engineering, TechnischeUniversitä  Carolo-Wilhelmina zuBraunschweig  International Center for Advanced Internet Research (iCAIR), Northwestern University  Korea Institute of Science and Technology Information (KISTI)  National Center for High-Performance Computing (NCHC) Taiwan  National Center for Supercomputing Applications (NCSA), University of Illinois  at Urbana-Champaign  NetherLight  Nortel  SARA  StarLight  SURFnet  Synchromedia  The BraunschweigUniversity of Art  University of Essex  University Van Amsterdam  18

HPDMnet Layer1 Topology 19

HPDMnet Layer2 Topology 20

Lessons Learned  Video transferred over FI testbed is not as smooth as over legacy Internet.  There are mosaics appearing every second.

 Because IGMP is not supported in OpenFlow, we have to manually insert multicast streaming flows into the flow table.

Future Work  Extend FI Testbed  Inter-OFCloud Control and Monitoring Development with Domestic Universities

Future Internet Testbed @ Taiwan OpenFlow Network OpenFlow Network at iCAIR /Chicago at NCHC OpenFlow Network (iGENI) at NCU TWAREN OpenFlow Network VPLS VPN at CHT-TL OpenFlow Network at NTUST OpenFlow Network Capsulator at NCKU @NTUST OpenFlow Network at KUAS

Monitoring on Multi-OFCloud OF Cloud OF Cloud @NCU @NTUST Controller Controller OF Cloud @NCKU Controller Monitoring Console OF Cloud Controller Controller @NCHC OF Cloud Data Plane @KUAS Manage Plane

Inter-Cloud Control and Monitoring  Each Cloud has its own OF Controller  Each Controller manages topology and flow provisioning inside the cloud  Inter-Cloud flow could be made by connecting partial flows provisioned by controllers of each cloud  Lack of global view for inter-cloud flows  No loops allowed for inter-cloud topology  Difficult to support QoS or SLA functions across clouds

Thank You 28