keith murphy supervisor: dr caspar ryan rmit university – distributed systems and networking...

Post on 17-Jan-2016

221 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Keith Murphy

Supervisor: Dr Caspar Ryan

RMIT University – Distributed Systems and NetworkingMelbourne, AUSTRALIA

Keith Murphy 1

Power Preservation Metrics Adaptation Algorithm Implementation

◦ Distributed Frameworks◦ MobJeX

Empirical Evaluation Expected Results Conclusion

Keith Murphy 2

Why is it important?◦ Mobile devices◦ Useability◦ More Advanced Applications

What has already been done?◦ Hardware monitoring of power.◦ Client – Server offloading.◦ Adaptation – Hardware and Software.

What Research is being Proposed?◦ Adaptation Algorithm for Distributed Frameworks.

Keith Murphy 3

Adaptation via transparent object migration within dynamic distributed networks.◦ useful for power preservation

JavaParty, Voyager, FarGo

MobJeX

Keith Murphy 4

What sort of metrics?◦ Metrics that have direct correlation to power

usage.

Identified Metrics:◦ Processor Utilisation◦ Network Utilisation◦ Memory Utilisation◦ Battery Status

◦ Direct versus Indirect measures (watts per unit).

Keith Murphy 5

Two Algorithms

Weighting algorithm that has been previously empirically evaluated in work by Pablo & Ryan.

An algorithm currently under development by Abebe & Ryan.

What is planned◦ Modify◦ Add new metrics

Keith Murphy 6

Designed to minimise the development effort .

Provides a middleware that detects changes within its environment via the collection of metrics.

automatically triggers the adaptation of applications via object migration using adaptation algorithms.

Different to the other frameworks where object migration is either predetermined or controlled by an administrator or scripts.

Predefined Metric collection and Adaptation Sub-systems.

Keith Murphy 7

Initial Simulation Scenario◦ Used to calibrate ◦ Comparing the following

Pablo & Ryan’s algorithm versus our enhanced power preservation version. Abebe & Ryan’s algorithm versus our enhanced power preservation version. Enhanced Pablo & Ryan’s algorithm versus Enhanced Abebe & Ryan’s.

◦ Testing: Different Resource Utilisation

Empirical Evaluation Scenario’s◦ Designed to Verify results from the simulations.

Keith Murphy 8

Power used whilst the algorithm completes a task will be measured.

The difference between the original and our enhanced power preservation version of it, will determine how much power has been saved.

This information will be graphed and an estimation of extended battery life of a said device will be made.

Keith Murphy 9

We expect see power being preserved

It will vary depending on the scenario

It is to early to estimate the power preserved

Keith Murphy 10

Feasible

At This Stage

Future Work

Questions?

Keith Murphy 11

Keith Murphy 12

top related