Space Application Solutions Using the Two Main Techniques

8/8/2019 Space Application Solutions Using the Two Main Techniques

1/22

-Harsh Vardhan

B.Tech-IT-V sem

SPACE APPLICATION SOLUTIONS USING THETWO MAIN TECHNIQUES OF SWARM COMPUTING


2/22

The greatest discoveries in the history of

mankind have been a result of mans

interaction with nature and vice versa- James T Elliot


3/22

INTRODUCTION

Increase in scale and complexity have given rise to the

need of development of reliable computing systems.

Fault tolerant systems are the need of the industry in

todays context.

Lately, self-managing systems has emerged as a similar

initiative that is being taken forward by autonomic

computing researchers.

Self-managing systems can be defined from two distinct

perspectives namely, the research perspective ofautonomic computing and the machine level perspective.


4/22

SELF MANAGING SYSTEMS


5/22

SIX MAIN TECHNIQUES OF

AUTONOMOUS COMPUTING


6/22

APPLICATION OF AUTONOMIC

COMPUTING TO SPACE APPLICATIONS

To achieve reliable systems , spacecrafts employ FPGAs

(Field Processing Gate Arrays) which are a special

purpose parallel processing system.

These FPGAs have a major drawback of being a subject toSEUs (Single Event Upsets).

These SEUs can occur due to radiation in outer space or

the sudden gravitational changes experienced in space.

These SEUs are the major points where the systems start

showing exceptional behaviour.


7/22

SOLUTIONS


8/22

IMPORTANT CONSIDERATIONS

If hardware is reconfigured then it would increase the

existing overheads on the system.

Hardware reconfiguration can be quite a challenging task

especially in zero gravity environments.

Replacement or servicing of hardware is a very limited

option in space.

Thus , software changes are easy to accomplish and can

produce magnificent results.


9/22

SWARM ARRAY + AGENT BASED

APPROACHES

A novel technique inspired by the Swarm robotic

techniques.

Actually, we integrate the swarm array and agent based

approaches to achieve autonomic computing.


10/22

SWARM ARRAY FRAMEWORK

Swarm array considers computational resources as aswarm of resources and the tasks to be executed as aswarm of tasks.

The idea is to consider complex interactions between

swarm of resources and swarm of tasks. The interactions between swarm agents bring about the

notion of intelligent agents or swarm agents carrying thesub-tasks and intelligent cores or swarm of coresexecuting the sub-task.

In other words, a computational approach emerging fromthe interaction of multi-dimensional arrays of swarmagents.


11/22

TWO APPROACHES TO SWARM

COMPUTING


12/22

APPROACH 1: INTELLIGENT CORES

The cores of a parallel computing system are considered

to be intelligent.

These intelligent cores are implemented as autonomous

swarm agents that represent the computing space.

A parallel task to be executed resides within a queue and

is scheduled onto the cores by a scheduler.

The intelligent cores interact with each other to transfer

tasks from one core to another at the event of a hardware

failure.


13/22

APPROACH 1: INTELLIGENT CORES

In this case, a processing core is similar to an organism

whose function is to execute a task.

The focus towards autonomy is laid on the parallel

computing cores abstracted onto intelligent cores.

The set of intelligent cores hence transform the parallel

computing system into an intelligent swarm.

To shift a process from one core to another, there is a

requirement of storing data associated and state of the

executing process, referred to as check pointing.


14/22


15/22

APPROACH 2: INTELLIGENT SWARMS

Only the task to be executed is considered to be intelligent

swarm agents that is autonomous.

A parallel task to be executed resides in a queue, which is

mapped onto carrier swarm agents by the scheduler.

The carrier swarm displace through the cores to find an

appropriate area to cluster and execute the task.

The intelligent agents interact with each other to achieve

mobility and successful execution of a task.


16/22

APPROACH 2: INTELLIGENT SWARMS

The task to be executed on the parallel computing cores

can be considered as a swarm of autonomous agents.

To achieve this, a single task needs to be decomposed and

the sub tasks need to be mapped onto swarm agents.

The swarm agents are only carriers of the sub-tasks or

are a wrapper around the sub-tasks.

The goal would be to find an area accessible to resources

required for executing the sub tasks within the

environment.


17/22


18/22

POINTS TO BE NOTED

The intelligent agents described above are an abstract

view of the sub-tasks to be executed on the hardware

cores.

The capabilities of the carrier swarm agents to identify

and move to the right location to execute a task.

In this case, the agents need to be aware of their

environments and which cores can execute the task.

The prediction of some type of core failures can be

inferred by consistent monitoring of power consumption

and heat dissipation of the cores.


19/22

SIMULATOR RESULTS FOR INTELLIGENT CORE

APPROACH IN EIGHT CONSECUTIVE TIME STEPS


20/22

SIMULATION RESULTS FOR AGENT BASED

APPROACH OVER EIGHT CONSECUTIVE TIME

STEPS


21/22

Almost all significant discoveries began with

a seemingly impossible speculation but

ended with a concrete result-Alexander Graham Bell


22/22

Documents

Space Application Solutions Using the Two Main Techniques