Vous êtes sur la page 1sur 4

Abstract:

offloading to cloud may improve performance of mobile applications but it is


not always the best bet because of complex Internet latency. Pytos a
framework for python offloading allows code migration to cloud-let
reducing Internet latency. Pytos provides a transparent off-loading
mechanism that automatically identifies the available cloud-lets. This paper
improves this offloading framework by giving more cloud-let options that are
IOT devices, mobile devices etc. so the offloading can be done to any of
these devices in that area hence increasing the availability of the cloud-lets.

Index Terms: mobile computing, Pytos: a python framework, offloading,


cloud-lets, mobile clustering.

Introduction:
wearable embedded devices has improved needs of technologies like
augmented reality, virtual reality, opinion mining etc. these operations are
costly in terms of computation. So these devices need high energy and
intensive computational power. Most users are unsatisfied with the battery
life of smart phones giving a poor rating of 6 out of 10. Initially mobile cloud
computing is introduced to solve such embedded devices and smart phone
challenges by proposing a distributed off-load mechanism in cloud. Here
computation are done in the high power machine in cloud. But this proposal
couldn't solve the real issue called Internet latency often called WAN delay.
As Pytos framework is local area network offloading mechanism along with
power reduction and performance boosting it also decreases the Internet
latency. But due to the limited cloud-let resources that can be available in a
local network in this paper we are proposing highly distributed mechanism
which indeed gives more strength to Pytos framework by introducing an
option called mobile clustering. So that there will be many devices which can
be acted as cloud-lets in the local network.

Implementation:
The proposed frameworks makes the offloading process for high
requirements computation with following steps that are: once Pytos identifies
methods of python to be offloaded, using the SSID it finds cloud-lets
available in the local network. If there are many devices and mobile phones
available then the framework distributes the task among the devices
connected in the network otherwise it performs the task locally. Pytos
daemon client takes care of new installations required to perform python
tasks in cloud-lets. Pytos evaluates whether offloading is an advantageous
task. If so tasks are offloaded other wise works are done in local machines.
Results of each cloud-lets are taken and combined then returned to the
machine that needs it.

The current evolution in mobile computing platforms follows the


developments in the desktop world, only at a much faster pace. Innovation
for mobile computing plat-forms is driven by a highly competitive market
and by the fast adoption of tablets which are situated between mobile
and desktop computing and therefore are always pursuing the highest
performance. Mobile computing platforms today surpass the computational
power of workstations from a few years ago. This combined with the fact that
desktop and server hardware is vastly outnumbered by mobile devices
deployed today, leads to the conclusion that we should find ways to fully
utilize these computational capacities. As there are millions of smart mobile
devices available it possible to have a high performance mobile clusters that
can be used for high end mobile computations. Mobiles are interconnected
via usb cables or wireless networks through usb hub and router.

There is a need for minimal setup to be able to make the mobile clustering on
smart phones, these access permission can only be obtained by rooting the
phone. To give all the relevant permissions and supports we used z4root tool.
Using debootstrap debian arm installations are done. The phone is rooted ato
get shell access and then chroot into the base Debian system installed as
mentioned. This method doesn't harm any existing softwares in the smart
phones. The mobile system can be functioned normally. The Debian userland
runs under the same kernel the phone has been booted with.

Pytos is development environment where programmers can choose how the


methods in the program are executed that is locally or in the cloud. This
framework checks whether the cloud-let in the local network are available, if
available then this python framework will do optimizing computations to
decide whether the runtime offloading is beneficial. Pytos library intercepts
the high intensive computation for a method. Pytos daemon clients takes care
of offloading decision making by optimistic computation. Pytos daemon
cloud-let is responsible for installing supportive softwares on network edges
or cloud-lets. Discovery end points are located at python web server gateway
interface intended to support client in the cloud-let discovery process.

In the above figure we have compared to three executional schemes local,


remote and collaborative execution in single cloud strategy with hetero
schema that is dynamic offloading strategy. Three observations are made that
are,the collaborative execution strategy is more flexible than the remote
scheme although it has higher energy consumption. Secondly,in the
heterogeneous mobile cloud, energy consumption and response time can be
reduced significantly based on the proposed offloading strategy, compared
with local execution and single resource condition. Above all, heterogeneity
of cloud resource and dynamic offloading strategy can efficiently improve
scalability of offloading service and energy efficiency of mobile device.

Vous aimerez peut-être aussi