IXP1200 Applications Ada Gavrilovska, Jiantao Kong, Weidong Shi, Xiaotong Zhuang Dr. Karsten Schwan,...
1
IXP1200 Applications IXP1200 Applications Ada Gavrilovska , Jiantao Kong, Weidong Shi, Xiaotong Zhuang Dr. Karsten Schwan, Dr. Ken Mackenzie Scalable Real Time Media Scalable Real Time Media Streaming on IXP1200 Streaming on IXP1200 Design Architecture Design Architecture DWCS on IXP1200 DWCS on IXP1200 Delta Air: Delta Air: An Operational Information An Operational Information System System Content-based processing in the Content-based processing in the IXP1200 IXP1200 IXP1200 as a cluster front-end, receives streaming data from external sources, and performs selective, application-specific mirroring and/or filtering, data striping or stream differentiation. RUDP used as underlying transport protocol at source nodes, IXPs and cluster destination- nodes, to guarantee reliable communication over the wide-area network. Motivation Combine IXP1200 and DWCS (dynamic window- constrained scheduling) as a media gateway to ensure QoS for large scale real time media streams IXP1200 Media Server Media Server Media Server Transm ition Level 0 segm ent0 segm ent1 segm ent99 Features of Our Design Multiple hierarchically indexed linear queues to speed up scheduling Less amount of SRAM and DRAM accesses per scheduling cycle than priority Queue Scheduling rate independent of number of active streams DWCS QoS Scheduler Sender thread Dropper thread Admission Control & Status Monitor StrongARM µEngines Protocol Processing & Enqueue Packet Receiver threads Packet Transmission Transmission threads SDRAM Stream Records Parallel packet scheduling and dropping so that packet dropping is not in the critical path of scheduling. High scheduling rate sustains thousands of streams Airport LAN Airport LAN Cluster Computing Real-Time Information Processing Simulation Optimization Baggage Status Operatio nal Flight Displays Baggage Displays Crew and Equipment Status FAA Flight Data Passenge r paging and response Gate Readers Wide-area Transport High Performance Computing Real-time Decision Tools Scalable Robust Services Storage capture, display, transport, filter, transform Real-Time Information Transport Visualization Recovery and Replay Real-time Situation Assessment Security Systems Equipment Inspection Airplane Data Traffic IXP Based Data Distribution Management IXP Based Data Distribution Management on Federated Simulation system on Federated Simulation system Example 2 Example 2 Example 1 Example 1 Example 3 Example 3 Data Distribution Management Federated Simulation Data Distribution Management Federated Simulation Data Distribution Management Federated Simulation Network Role of Data Distribution Management in Distributed Simulation Filtering federates should only receive data that they want Group Management entity data subscribe and publish. Each Federate simulates certain number of entities. Data Distribution Management Federated Simulation IXP IXP IXP Network Federates Federates Federates Dynamic application interest management. Each federate should only receive the data it wants. Multicast and group management. Simulation often requires large number of groups. The system should scale well when the number of groups increases. Fast delivery of object states and attributes. Solution for DDM on IXP1200 Solution for DDM on IXP1200 Active Interest filtering group management and multicast Simulation data cache Challenges in DDM Challenges in DDM Application-level Filtering and Application-level Filtering and Mirroring Mirroring RUDP necessary since application- specific handlers operate on data content, not just header information. Data formats are used to interpret bits in memory and access desired data fields content. Handlers can be invoked as soon as necessary data present, which can prevent potentially unnecessary memory accesses. For smart mirroring output data can be customized in a destination-specific manner
-
Upload
christal-gibbs -
Category
Documents
-
view
231 -
download
0
Transcript of IXP1200 Applications Ada Gavrilovska, Jiantao Kong, Weidong Shi, Xiaotong Zhuang Dr. Karsten Schwan,...
IXP1200 ApplicationsIXP1200 ApplicationsAda Gavrilovska , Jiantao Kong, Weidong Shi, Xiaotong Zhuang
Dr. Karsten Schwan, Dr. Ken Mackenzie
Scalable Real Time Media Streaming on Scalable Real Time Media Streaming on IXP1200IXP1200 Design ArchitectureDesign Architecture DWCS on IXP1200DWCS on IXP1200
Delta Air: Delta Air: An Operational Information An Operational Information SystemSystem Content-based processing in the IXP1200Content-based processing in the IXP1200
IXP1200 as a cluster front-end, receives streaming data from external sources, and performs selective, application-specific mirroring and/or filtering, data striping or stream differentiation.
RUDP used as underlying transport protocol at source nodes, IXPs and cluster destination-nodes, to guarantee reliable communication over the wide-area network.
Motivation
Combine IXP1200 and DWCS (dynamic window-constrained scheduling) as a media gateway to ensure QoS for large scale real time media streams
IXP1200
Media Server
Media Server
Media Server
Transmitionpointer
Level 0
segment 0 segment 1 segment 99
Features of Our Design Multiple hierarchically indexed linear queues to speed
up scheduling Less amount of SRAM and DRAM accesses per
scheduling cycle than priority Queue Scheduling rate independent of number of active
streams
DWCS QoS SchedulerSender thread Dropper thread
DWCS QoS SchedulerSender thread Dropper thread
Admission Control& Status MonitorAdmission Control& Status MonitorStrongARM
µEngines
Protocol Processing &Enqueue PacketReceiver threads
Protocol Processing &Enqueue PacketReceiver threads
Packet TransmissionTransmission threadsPacket TransmissionTransmission threads
SDRAMStream Records
SDRAMStream Records
Parallel packet scheduling and dropping so that packet dropping is not in the critical path of scheduling.
High scheduling rate sustains thousands of streams
AirportLAN
AirportLAN
AirportLAN
AirportLAN
Cluster ComputingReal-Time
Information Processing
Cluster ComputingReal-Time
Information Processing
Simulation Optimization
Baggage Status
OperationalFlight Displays
Baggage Displays
Crew andEquipmentStatus
FAA FlightData
Passenger paging andresponse
GateReaders
Wide-areaTransport
High Performance ComputingReal-time Decision Tools
Scalable Robust Services
Storage
capture, display, transport, filter, transform
Real-Time Information Transport
Visualization
Recovery and Replay
Real-timeSituation
Assessment
Security Systems
EquipmentInspection
AirplaneData Traffic
IXP Based Data Distribution Management on IXP Based Data Distribution Management on Federated Simulation systemFederated Simulation system
Example 2Example 2
Example 1Example 1
Example 3Example 3
Data Distribution Management
Federated Simulation
Data Distribution Management
Federated Simulation
Data Distribution Management
Federated Simulation
Network
Role of Data Distribution Management in Distributed Simulation
Filtering federates should only receive data that they wantGroup Management entity data subscribe and publish.
Each Federate simulatescertain number of entities.
Data Distribution Management
Federated Simulation
IXP
IXP
IXP
Network
Federates
Federates
Federates
Dynamic application interest management.Each federate should only receive the data it wants.
Multicast and group management. Simulation often requires large number of groups. The system should scale well when the number of groups increases.
Fast delivery of object states and attributes.
Solution for DDM on IXP1200Solution for DDM on IXP1200
Active Interest filteringgroup management
and multicastSimulation data cache
Challenges in DDMChallenges in DDM
Application-level Filtering and MirroringApplication-level Filtering and Mirroring
RUDP necessary since application-specific handlers operate on data content, not just header information.
Data formats are used to interpret bits in memory and access desired data fields content.
Handlers can be invoked as soon as necessary data present, which can prevent potentially unnecessary memory accesses.
For smart mirroring output data can be customized in a destination-specific manner
![MutS HOMOLOG1-Derived Epigenetic Breeding · MutS HOMOLOG1-Derived Epigenetic Breeding Potential in Tomato1[OPEN] Xiaodong Yang, Hardik Kundariya, Ying-Zhi Xu, Ajay Sandhu, Jiantao](https://static.fdocuments.net/doc/165x107/5ff79586b9e48e2a58283ba9/muts-homolog1-derived-epigenetic-muts-homolog1-derived-epigenetic-breeding-potential.jpg)
![Enabling Fine-Grained Edge Offloading for IoThome.in.tum.de/~ding/files/sigcomm-poster2017.pdf[1] Ketan Bhardwaj, Ming-Wei Shih, Pragya Agarwal, Ada Gavrilovska, Taesoo Kim, and Karsten](https://static.fdocuments.net/doc/165x107/5fdbe9bc901d59681b1242d5/enabling-fine-grained-edge-offloading-for-dingfilessigcomm-poster2017pdf-1.jpg)
