Presentation is loading. Please wait.

Presentation is loading. Please wait.

Cooperative Computing: A Computing Model for Large Networks of Embedded Systems Cristian Borcea, Phillip Stanley-Marbell, Kiran Nagaraja, Liviu Iftode.

Similar presentations


Presentation on theme: "Cooperative Computing: A Computing Model for Large Networks of Embedded Systems Cristian Borcea, Phillip Stanley-Marbell, Kiran Nagaraja, Liviu Iftode."— Presentation transcript:

1 Cooperative Computing: A Computing Model for Large Networks of Embedded Systems Cristian Borcea, Phillip Stanley-Marbell, Kiran Nagaraja, Liviu Iftode Rutgers University

2 Networks of Embedded Systems (NES) Today  Characteristics  Limited resources  Large scale  Heterogeneous  Volatile  Unattended  Example: Networks of Sensors  Applications: data collection/dissemination  Research: ad-hoc routing

3 Next Generation NES  More powerful systems ( processor, memory, network )  More demanding applications ( object tracking )  How to execute user-defined distributed applications ?  computing model  system architecture

4 Traditional Distributed Computing  Assumptions  functionally homogeneous nodes  assumes stable configuration  fixed addressing scheme  exact results  Inadequate for NES

5 The Cooperative Computing Model  Distributed computation on large scale ad-hoc networks  The set of nodes involved in computation:  identified by their properties  discovered using application controlled routing  Partial execution acceptable when a certain Quality of Result ( QoR ) is met

6 Application Example  Compute the average temperature over red nodes  QoR: average over at least 3 red nodes  Red nodes used for computation  Blue and green nodes used as intermediate hops

7 System Architecture  Smart Messages ( SMs )  migrate through the network searching for target nodes  execute on each node  Minimal System Support  admission  scheduling and execution  synchronization  communication, but no routing

8 Smart Messages SignatureResource TableCode BricksData Bricks  Code and data bricks  Signature-based authentication for access control  Resource table: estimated resource requirements for admission control

9 System Support Operating System Hardware Tag Space Virtual Machine Admission Manager

10 Tag Space  Tasks create, delete, read, write tags  Tags discarded when lifetime expires.................. Temperature QxwyZ 7200 80 Name Signature Lifetime Data

11 Admission  At arrival SM presents its resource requirements  tags to be created/accessed  estimated memory requirements, execution time, network traffic  Each admitted SM generates a new task

12 Execution Tag Space red tag ? Computation: sum=5+2 sum=5 sum=7 c2c1d1d2c1c2d1d2 c1 c2 d1 d2 Task1 Node red=2

13 Scheduling and Synchronization  FIFO scheduling  Non-preemptive execution with resource protection  Update-based synchronization on tags

14 Self-Routing  Routing done entirely by the application  SMs carry code and data for routing  Tag space stores signed routing information

15 Self-Routing Example Network Red tag ? rRedval c1c2d1d2 c1 d1 c2 d2 c2 d2 rRed Task1 Spy SM sent Spy SM returns Task2 c2d2 c2d2 c1c2d1d2 Tag Space Tag Space Node iNode j

16 Prototype Software Infrastructure uClinux Admission Manager Modified KVM Message Queue Tag Space Temporary Receive SM Send SM uCsimm & Bluetooth

17 Contributions  Cooperative Computing: distributed computing model for networks of embedded systems  System architecture: Smart Messages  active carriers of data  integrate computation and communication  application controlled routing ( self-routing )  http://discolab.rutgers.edu/projects/sm.htm  Accepted as position summary at HotOS VIII

18 Future Work  Evaluate the tradeoffs between flexibility and overhead of migration  Define a partially successful execution  Enforce more complex security policies  Simulate various applications and routing algorithms  Integrate energy in the model


Download ppt "Cooperative Computing: A Computing Model for Large Networks of Embedded Systems Cristian Borcea, Phillip Stanley-Marbell, Kiran Nagaraja, Liviu Iftode."

Similar presentations


Ads by Google