Scalability of Vehicle Networks through Vehicle Virtualization
Jiangchuan Huang, Christoph M. Kirsch, Raja Sengupta

Citation
Jiangchuan Huang, Christoph M. Kirsch, Raja Sengupta. "Scalability of Vehicle Networks through Vehicle Virtualization". Talk or presentation, 29, September, 2013; Poster from the First International Workshop on the Swarm at the Edge of the Cloud (SEC'13 @ ESWeek), Montreal.

Abstract
We explore extending the paradigm of cloud computing to computing tasks having locations in space. A computing task is a triple (ArrivalTime, Location, ComputingTime). A server is considered to execute such a task by visiting the task Location and staying there for ComputingTime, at any time after the task ArrivalTime. Sampling applications in time and space, such as those entailed by Google street view, mobile sensor networks, or real-time traffic reporting by radio stations, are examples of this type of computing task. The servers, henceforth referred to as real vehicles, are networked vehicles each having all or some of the sensing, computation, communication, and locomotion capabilities. Examples include robots like driver-less cars, drones, or manned helicopters, traveling and then pausing to execute computing tasks like taking pictures, measurements or monitoring at specified places. We organize the collection of moving servers as a new type of cloud called the spatial cloud, or the Cyber- Physical Cloud (CPC). CPC may work better than conventional cloud computing because of migration gain, a concept unique to the spatial cloud introduced below. We extend the idea of the virtual machine used in cloud computing to an idea called the virtual vehicle to create performance isolation. Just as the cloud computing customer has a service-level agreement (SLA) for a virtual machine, our customer has an SLA for a virtual vehicle. To a customer, a virtual vehicle is exactly like a real vehicle that travels at a virtual speed specified in the SLA. For example, suppose a radio station (the customer) uses a helicopter to overfly accident scenes (the task) at 20 mile per hour (mph) for real-time traffic reporting. Such a helicopter would arrive at an accident scene 2 miles away in 6 minutes. We now virtualize this helicopter. Instead of buying or renting, and operating a real helicopter, the radio station would buy virtual helicopter that travels at 20 mph using an SLA. Our cyber-physical cloud would then take responsibility for flying some real helicopter to the accident site in 6 minutes or some near approximation to this time. Our results show that small deviations from the 6 minutes can turn into large numbers of virtual vehicles being realized by many fewer real vehicles. This is the cyber-physical cloud computing gain. The radio station could enjoy large reductions in cost by tolerating small variations in the 6 minutes. The real vehicles travel to and execute each task such that the real completion time is no later than the expected completion time. The system achieves high performance isolation if a statistically dominant subset, e.g., 98%, of the virtual vehicle’s tasks are completed no later than their expected completion time such that the customer does not even notice a small overflow over expected completion times. The expected completion time behaves like a “deadline” to the provider. We call it the virtual deadline. The virtual deadlines make the spatial cloud a soft real-time system. We use performance metrics such as tardiness and delivery probability to measure performance isolation. How the provider realizes the virtual deadlines of the tasks is of no concern to the customer. For example, the provider can use a real vehicle to travel to and execute a task as the virtual vehicle does, or migrate the information of the virtual vehicle through a network to another real vehicle closer to the task. In the latter case, the real distance traveled is smaller than the virtual distance resulting in the phenomenon we have called migration gain. The idea works when the communication costs of migrating a virtual machine over the network are small. The theorems and simulation results show that the provider can support a given number of virtual vehicles with significantly fewer real vehicles while guaranteeing high performance isolation. We quantify the gain by the ratio of the number of virtual vehicles over the number of real vehicles. The gain arises from two phenomena. (i) A customer may not fully utilize her virtual vehicle, enabling the spatial cloud to multiplex several virtual vehicles onto one real vehicle. This type of gain is called multiplexing gain. It is known in communication networks and cloud computing. (ii) The real vehicles save travel distance by migrating the virtual vehicle hosting a task to another real vehicle closer to the task, creating migration gain. Our results focus on migration gain since it is unique to the spatial cloud. Migration gain is the reason spatial cloud computing outperforms conventional cloud computing.

Electronic downloads

Citation formats  
  • HTML 
    Jiangchuan Huang, Christoph M. Kirsch, Raja Sengupta. <a
href="http://www.terraswarm.org/pubs/137.html"><i>Scalability
of Vehicle Networks through Vehicle
Virtualization</i></a>, Talk or presentation, 
29, September, 2013; Poster from the <a
href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
>First International Workshop on the Swarm at the Edge of
the Cloud (SEC'13 @ ESWeek)</a>, Montreal.
  • Plain text 
    Jiangchuan Huang, Christoph M. Kirsch, Raja Sengupta.
"Scalability of Vehicle Networks through Vehicle
Virtualization". Talk or presentation,  29, September,
2013; Poster from the <a
href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
>First International Workshop on the Swarm at the Edge of
the Cloud (SEC'13 @ ESWeek)</a>, Montreal.
  • BibTeX 
    @presentation{HuangKirschSengupta13_ScalabilityOfVehicleNetworksThroughVehicleVirtualization,
    author = {Jiangchuan Huang and Christoph M. Kirsch and Raja
              Sengupta},
    title = {Scalability of Vehicle Networks through Vehicle
              Virtualization},
    day = {29},
    month = {September},
    year = {2013},
    note = {Poster from the <a
              href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
              >First International Workshop on the Swarm at the
              Edge of the Cloud (SEC'13 @ ESWeek)</a>, Montreal.},
    abstract = {We explore extending the paradigm of cloud
              computing to computing tasks having locations in
              space. A computing task is a triple (ArrivalTime,
              Location, ComputingTime). A server is considered
              to execute such a task by visiting the task
              Location and staying there for ComputingTime, at
              any time after the task ArrivalTime. Sampling
              applications in time and space, such as those
              entailed by Google street view, mobile sensor
              networks, or real-time traffic reporting by radio
              stations, are examples of this type of computing
              task. The servers, henceforth referred to as real
              vehicles, are networked vehicles each having all
              or some of the sensing, computation,
              communication, and locomotion capabilities.
              Examples include robots like driver-less cars,
              drones, or manned helicopters, traveling and then
              pausing to execute computing tasks like taking
              pictures, measurements or monitoring at specified
              places. We organize the collection of moving
              servers as a new type of cloud called the spatial
              cloud, or the Cyber- Physical Cloud (CPC). CPC may
              work better than conventional cloud computing
              because of migration gain, a concept unique to the
              spatial cloud introduced below. We extend the idea
              of the virtual machine used in cloud computing to
              an idea called the virtual vehicle to create
              performance isolation. Just as the cloud computing
              customer has a service-level agreement (SLA) for a
              virtual machine, our customer has an SLA for a
              virtual vehicle. To a customer, a virtual vehicle
              is exactly like a real vehicle that travels at a
              virtual speed specified in the SLA. For example,
              suppose a radio station (the customer) uses a
              helicopter to overfly accident scenes (the task)
              at 20 mile per hour (mph) for real-time traffic
              reporting. Such a helicopter would arrive at an
              accident scene 2 miles away in 6 minutes. We now
              virtualize this helicopter. Instead of buying or
              renting, and operating a real helicopter, the
              radio station would buy virtual helicopter that
              travels at 20 mph using an SLA. Our cyber-physical
              cloud would then take responsibility for flying
              some real helicopter to the accident site in 6
              minutes or some near approximation to this time.
              Our results show that small deviations from the 6
              minutes can turn into large numbers of virtual
              vehicles being realized by many fewer real
              vehicles. This is the cyber-physical cloud
              computing gain. The radio station could enjoy
              large reductions in cost by tolerating small
              variations in the 6 minutes. The real vehicles
              travel to and execute each task such that the real
              completion time is no later than the expected
              completion time. The system achieves high
              performance isolation if a statistically dominant
              subset, e.g., 98%, of the virtual vehicleâs
              tasks are completed no later than their expected
              completion time such that the customer does not
              even notice a small overflow over expected
              completion times. The expected completion time
              behaves like a âdeadlineâ to the provider. We
              call it the virtual deadline. The virtual
              deadlines make the spatial cloud a soft real-time
              system. We use performance metrics such as
              tardiness and delivery probability to measure
              performance isolation. How the provider realizes
              the virtual deadlines of the tasks is of no
              concern to the customer. For example, the provider
              can use a real vehicle to travel to and execute a
              task as the virtual vehicle does, or migrate the
              information of the virtual vehicle through a
              network to another real vehicle closer to the
              task. In the latter case, the real distance
              traveled is smaller than the virtual distance
              resulting in the phenomenon we have called
              migration gain. The idea works when the
              communication costs of migrating a virtual machine
              over the network are small. The theorems and
              simulation results show that the provider can
              support a given number of virtual vehicles with
              significantly fewer real vehicles while
              guaranteeing high performance isolation. We
              quantify the gain by the ratio of the number of
              virtual vehicles over the number of real vehicles.
              The gain arises from two phenomena. (i) A customer
              may not fully utilize her virtual vehicle,
              enabling the spatial cloud to multiplex several
              virtual vehicles onto one real vehicle. This type
              of gain is called multiplexing gain. It is known
              in communication networks and cloud computing.
              (ii) The real vehicles save travel distance by
              migrating the virtual vehicle hosting a task to
              another real vehicle closer to the task, creating
              migration gain. Our results focus on migration
              gain since it is unique to the spatial cloud.
              Migration gain is the reason spatial cloud
              computing outperforms conventional cloud computing.},
    URL = {http://terraswarm.org/pubs/137.html}
}

Posted by Christopher Brooks on 2 Oct 2013.

Notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright.