Information Technology Reference
In-Depth Information
For the second and the third issue, we need to implement the smart phone
application and the server ourselves. At this moment, we know which layer's
trac the operators are charging for, so we can deliberately let the client send
specific protocol packets to the server, and observe whether the operators will
charge for this trac, and then we can get the result of the second issue. We
can let the client send data as little as possible using the protocol that will be
charged, and see how much the data volume recorded by the operators. Thus
get the result of the third issue.
To make clear of the fourth issue, we just need to know whether the operators
are using the NAT (Network Address Translator) technology, if so, what kind of
NAT are they using? Only the Port Restricted Cone and Symmetric NAT have
no such loophole [6]. If the address of the device is an internal IP, we know that
the operator is using NAT. If so, we deploy two servers with different IPs running
same service, let the smart phone application uses a same port to send packets to
the two servers. Then the servers can obtain the source port of received packets.
If the source ports are the same at the two servers, the operator is using Cone
NAT, otherwise Symmetric NAT [6].
There are two methods to obtain data usage logged by operators. The first
one is via the SMS from the operators, the second is to login the operators
website and inquire data usage. China Mobile and China Telecom support both
methods. But for getting finer result, we choose the second method for both
of them. China Unicom SIM card used in our experiment only support the first
method, the device will receive the billing SMS only when its trac consumption
since last billing SMS is greater than 50 KB.
After exploring the four aspect of the operators' charging system, we can fur-
ther take evaluation of tra
c consumption of smart phone applications. Because
users and developers only concern tra
c consumption charged by the operators,
we could get more accurate comparison if we just taking this part of trac into
account. Table 1 lists the application kinds and corresponding applications to
be evaluated. They are all content controllable applications. For the first issue,
we take experiment in different scenarios on each type of applications, and use
TraSt to get the part of trac that is generated by the application and will be
charged, and then we can compare the result among the applications. For the
second issue, we can further analysis the trac with the specific application, and
try to get the root cause. Notice that we are evaluating content controllable ap-
plications. For the applications of the same type, we can make the data content
basically the same at each network condition repeatedly.
Our mobile devices are HTC Desire and HTC Incredible S, running on An-
droid 2.2 and 2.3.1 respectively.
5Rsut
In this section, we will show our testing result for the operators' charging system
and evaluation results for the four kinds of applications. All the following results
are obtained in May 2012.
