Introduction to Internet of Things

Prof. Sudip Misra

Department of Computer Science and Engineering
Indian Institute of Technology, Kharagpur

Lecture - 36
Software - Defined IoT Networking - Part- II

So, now we are going to discuss about software defined networking and its use in IoT.
So, here we are going to talk about how SDN can help in improving mobile networks
specifically.

(Refer Slide Time: 00:39)

So, what we have already discussed in the first part of SDN for IoT lecture is that we
have. So, a sensor networks and software defined networking can be used to improve the
performance of sensor networks. So, we have different SDN approaches for WSN and
different approaches like Sensor OpenFlow, Soft-WSN, SDN-WISE are already
available in the literature which can be used. Community can use it if they want to
implement sensor you know SDN for WSN and we have also discussed about how
software defined WSN can improve the overall performance of the network compared to
the traditional WSN.
(Refer Slide Time: 01:40)

So, this is what we discussed earlier. Now, let us talk about the Traditional Mobile
Wireless Networks. There are different problems in these Traditional Mobile Networks
and these networks can be like Wi-Fi based networks. Mobile cellular networks or other
mobile networks, the problems are with respect to scalability. So, you know what
happens is each of these nodes in the networks if the nodes including the base stations
access points etcetera, typically they are statically over provisioned you know. So, there
is over provisioning in each of the nodes in these networks and they are inflexible to
manage the mobile traffic with high demand.

They are difficult to manage many times. This leads to misconfigurations. They are
inflexible and requires too much time to introduce a new service as a hardware
architecture is inaccessible and they are cost expensive, both in capital expenditure and
operational expenditure terms because both the capex as well as the opex are very
expensive.
(Refer Slide Time: 02:55)

So, two things are important and this is this information has been adopted from the Open
Network Foundation. So, two things are important when we want to adopt SDN for
mobile networks. One is the flow table paradigm of SDN which is well suited for end to
end communication over multiple technologies, such as Wi-Fi, 3G, 4G and so on so
forth. The other one is the logically centralized control which is particularly useful for
efficient base station, coordination for addressing intercellular interference. So, we are
going to talk about each of these in slightly more detail in the next little while.

(Refer Slide Time: 03:41)

Other issues which are also important are path management which concerns basically
data that can be routed based on service requirements without depending on the core
routing policies and network virtualization which is about abstracting the physical
resources from the network services. This helps in providing seamless connectivity and
service differentiation among the users.

(Refer Slide Time: 04:07)

So, in a traditional mobile network, what happens is when a mobile equipment like this
operates, it connects with a base station or an axis point for regular data. You know data
back and forth between it and that access point and this particular node, it serves this
particular access point, sorry this particular user equipment, but while it is doing, it is
also getting, it may also get affected by interference from another node. So, this red
colored arrows basically shows the interference and the blue colored ones basically
shows the regular data, transfer data communication.

So, there is this solution which is this SDWMN. It tries to address this particular problem
of interference through the help of SDN. So, here what we see is the architecture of the
software defined mobile network and as we can see over here, we have an SDN
controller and Virtualized Radio Resource Management Unit which basically takes care
of interference with the problem of interference. So, this particular unit what it does is, it
takes care of issues such as the signals of ENodeB 2 and the signal of ENodeB 3, they do
not affect each other. That means, the signal of ENodeB 2 will not affect the signal of
ENodeB 3 affect means through interference.

So, this is made possible through this Virtualized Radio Resource Management Unit
which has been proposed in this particular protocol or the solution which is SDWMN
and with respect to mobile traffic management.

(Refer Slide Time: 06:27)

There is this unit which is the ANDSF, Access Network Discovery and Service Function
Unit which basically takes care of it and this is done this way. So, we have the cellular as
well as the Wi-Fi and the mobility. So, basically the what happens is when the user is
moving, so the user can get first connected to the regular. You know Wi-Fi access point,
then you know it might have better connectivity with the cellular network 3G, 4G and so
on.

So, basically this particular solution basically takes care of it how to interoperate
between these different when the user is moving and is getting connected with you know
and has different signals from different devices from different networks, Wi-Fi network,
3G, 4G networks etcetera. How it is going to do this switching? So, the SDN basically
comes to a rescue to solve this particular problem.
(Refer Slide Time: 07:35)

The key benefits of the solution is that centralized control of devices manufacture by
multiple vendors is made possible. It is possible to have higher rate of integration of new
services and abstracted network control and management is made possible through this
particular solution with that option of SDN in these mobile networks.

(Refer Slide Time: 08:01)

So, you know when we talk about SDN and in SDN based solution, it is about rules, flow
rules and rule placement controller placement and so on. So, let us now look at how rule
placement is made at the access devices in these networks. The challenges that have to be
addressed are that number one, the general open flow does not support wireless network.
So, you know because it is general open flow does not do that. Some kind of a
modification of the existing open flow is required. Number two, typically the users are
mobile in nature and the network is highly dynamic. Third challenge that has to be
addressed with respect to rule placement in these networks is that the frequent changes in
rule placement is required and the presence of heterogeneous devices in the network
might is a reality in these scenarios.

So, basically you know what happened at that last point, what it says the last challenge
what it says is different types of heterogeneous devices having different configurations
supporting different you know vendor specific solutions and protocols and so on exist in
a realistic mobile based IoT network. So, how to support such heterogeneous devices in a
single platform? So, these are the challenges that have to be worked upon in order to deal
with the rule placement issue.

(Refer Slide Time: 09:36)

So, there are different approaches or different solutions that have been proposed for it.
Number one is Odin, number two is Ubi-Flow and number three is Mobi-Flow. Odin
basically proposes the use of an agent. It is an agent based solution, agent based
architecture.
(Refer Slide Time: 09:48)

So, here an agent is placed at the access points to communicate with the controller. So,
this is the Odin agent and there is the Odin master. So, we have the Odin agent which is
placed at the physical devices at the access point. For example, the Odin master which is
placed at the controller and now this is this hand shaking between the communication
between the Odin agent and the Odin master that is taken care of by this Odin protocol.

(Refer Slide Time: 10:24)

So, it uses the communication between the Odin agent and the Odin master. The
corresponding flow diagram, the flowchart is shown over here. We do not need to
understand in detail, but it is given for those who might be interested to understand it
nonetheless and so, basically it is a handshaking that goes on between this Odin agent,
add the access device and access point and the Odin master at the controller that makes
Odin function.

(Refer Slide Time: 10:57)

Now, let us look at the second solution, the Ubi-Flow solution which was published in
infocom very recently in 2015 only, 2 years back.

So, here there are two main parts of the solution. One takes care of Mobility
Management and the other one takes care of Flow Scheduling. Mobility management
takes care of issues such as you know how to have scalable control of the access points
issues, such as fault tolerance. That means, if some component goes down, then what
happens you know how can the network still function and the other one is the flow
scheduling which has three parts. First is partitioning this entire network into multiple
partitions, no fragmenting this entire network into multiple partitions and over here we
have shown those three partitions through three different color shades i.e. grey, light
green and orange.

So, we have three different partitions shown over here network partitions and after this
network partitioning has happened, then you know network matching has to be done. So,
let us say that this particular mobile device receives a particular packet. So, then what it
is going to do? It is going to perform, it is going to execute this network matching,
network matching algorithm. What that network matching algorithm is going to do is, it
is going to check that which of these network partitions should handle this particular
packet.

So, then it will be sent to that particular partition for further handling and load balancing
basically takes care of it you know. So, you know how the network matching and load
balancing will be done together. So, the load balancing is going to take care of issues,
such as that one partition should not get too much loaded while the other partition is
under loaded and so on. So, this is what the load balancing is going to do.

(Refer Slide Time: 13:09)

Now, let us look at the third solution that I was talking about the Mobi-Flow and this
paper was published in IEEE Globecom, 2016 and this is the solution Mobi-Flow
solution was proposed by us in the Swain group and Mobility. So, Mobi-Flow basically
gives a mobility aware flow, flow rule placement for mobile based IoT solutions, mobile
place based IoT systems. So, let us look at these three scenarios over here. So, here we
have different nodes S1 M1 connected to this particular access point. M2 and P1
connecting to another access point and their corresponding flow tables are shown here,
the corresponding flows tables corresponding to this particular access point and this
particular access point of the base station, the corresponding flow table is given over
here. It is very easy to understand this part.
Now, let us see, let us say that the nodes M1 and P1 have interchanged their positions.
Let us assume a very simple scenario, where M1 and P1 have interchanged their
positions and the current situation looks like this. So, essentially you know what should
happen is, we will have something like this. The another flow rule has to be added in the
flow table corresponding to this particular access point and the base station in this
particular manner, but at the same time we also know that flow table spaces are very
limited.

So, we cannot simply have these flow tables grow with the addition of these flow rules.
So, you know what essentially has to be done is the scenario like this. So, how can we do
it, this is what Mobi-Flow talks about.

(Refer Slide Time: 15:02

So, Mobi-Flow gives a proactive rule placement scheme depending on the users
movement in the network and when we talk about such a scheme, you know what is very
important to have some kind of prediction of the location of the end users at the next
time instant give you the data of how the user was moving till a particular time instant.
(Refer Slide Time: 15:36)

Then, placing the flow rules at the access points which can be associated to the users
based on the predicted locations, location prediction, this is done in Mobi-Flow using
something known as the Order-K Markov Predictor which basically takes the last k th
location instances to predict the next location and the flow rule placement which
basically is a linear programming based solution that is used to select which access point
is going to be the optimal one.

(Refer Slide Time: 16:01)

So, for further details the difference of Mobi-Flow was given in 2016 Globecom paper
and if somebody wants to have more insight about how Mobi-Flow works, so one needs
to go through that particular paper, but I would like to summarize by giving a summary
of the results of comparison between Mobi-Flow and the conventional solution.

So, what we have is with respect to message overhead and energy consumption, Mobi-
Flow performs better compared to the conventional network, conventional solution and
this is quite evident from these two plots. So, control message overhead and energy
consumption can be minimized significantly using Mobi-Flow compared to the
conventional tool placement schemes.

(Refer Slide Time: 16:59)

So, next we have to take care of how to perform rule placement at the backbone network.
So, the existing rule placement schemes for wired networks can be used over here
because most of the backbone networks have a topology and the structure. The overall
architecture is similar to that which exists for wired networks.

So, we can use the existing rule placement schemes for wired networks over here as well
for the backbone network of IoT and the load balancing is an important issue due to the
dynamic nature of the IoT network. So, dynamic resource allocation can also be
integrated.
(Refer Slide Time: 17:39)

So, another very important thing that we have to remember in this particular context is
the Data Center Networking. So, there are two types of flows that occur in Data Center
Networks. So, Data Center Networks basically you know here we are talking about
implementation of data center networks with the help of SDN.

So, typically two different solutions of adoption of SDN are in data center networks. So,
here we have typically two types of flows that are observed. One is the Mice-Flow which
are basically you know small flows and the other one is Elephant-Flow which is
basically you know large scale flows where big volumes of data are coming, big sized
data are coming and so on and small flows, Mice-Flows where small sized data are
coming. So, here what is suggested is for the Mice-Flow wildcard rules can be placed to
deal with these flows, the wildcard rules and for the Elephant-Flows what is required is
to have exact match of the rules.

So, we need to classify the flows before inserting the flow rules at the switches to
adequately forward them in the network. So, if it is an elephant flow, you know we have
to have exact match. If it is a Mice-Flow, we will go for a wildcard match. Anomaly
detection can also be performed in IoT network using SDN or open flow.
(Refer Slide Time: 19:07)

So, here you know it is required to monitor the network through open flow to detect any
anomaly in the network which can be done by monitoring each flow in the network. It is
also possible to collect different port statistics at the different switches and thereby from
these statistics, you know anomaly detection techniques can be implemented and
different anomalies can be found out.

(Refer Slide Time: 19:38)

Now, what is required is to experiment with these wireless networks. So, what are the
different platform that are available? So, Mininet traditionally did not have any wireless
support, but recently there is this Mininet-WiFi that is available which can be used to
deploy a network support. Mininet-WiFi can help support both wired and wireless
network. Wired implements the Ethernet protocol and Wireless implements 802.11 class
of protocols. So, this is one platform that is available. Mininet-WiFi, the other one is the
ONOS platform which can be used to place the controllers. ONOS basically helps in
placement of the controller.

(Refer Slide Time: 20:21)

So, it helps to solve the controller placement problem. So, again to sum up SDN based
solutions are useful to deal with different challenges present in an IoT network and in the
first part of it, we talked about IoT in the context of sensor networks and small power,
you know small devices, small IoT devices and in the second part, we talked about IoT in
the context of mobile networks, mobile devices and so on and we have talked about three
different solutions. One is the Ubi-Flow, Mobi-Flow and Odin. These are the three
different solutions that are available recently to support SDN for mobile IoT networks.

So, there are different issues of interference and mobility management which have to be
taken care of in these SDN based approaches to SDN based approaches to IoT. So,
interference is a very common thing, a very important problem that has to be addressed
and mobility as well because mobility is something that occurs in any IoT network and
particularly the mobile based IoT networks.
So, with this we come to an end and we will stop here. We have understood that how
SDN can be used for implementing IoT networks and making the IoT networks much
more efficient.

Thank you.