Transcript Supporting the Brave New World of the 4As: Anytime, Anywhere

Supporting the Brave New World of
the 4As: Anytime, Anywhere,
Anyhow and Anything
Glenford Mapp
Associate Professor
Middlesex University
United Kingdom
NOVEMBER 2013
WTD-ICMC-USP
Outline of My Talk

What is this new world really about

What are the challenges

Y-Comm – a bridge to this new world

Collaborations

What's new in my world

Joining the revolution
NOVEMBER 2013
WTD-ICMC-USP
Anytime, Anywhere

Anytime


It's a generational thing
Anywhere

Actually I want to be able to communicate from
anywhere in the universe.
•
•
NOVEMBER 2013
So I will need to spawn networks and connect
them to existing networks
Building, managing and controlling networks
should be under user control
WTD-ICMC-USP
Anyhow

We are seeing the rise of new networks based
on different technologies
•

Wi-Fi, WiMax, 4G, Ultrawideband, Optical
Networks, etc
We are also seeing new types of networks
•
•
•
•
•
•
NOVEMBER 2013
Delay Tolerant Networks (DTNs)
Home Networks, Personal Area Networks,
VANETs – Vehicular networks
Infrastructural networks: Sensor Networks
E-Health Networks: Patient monitoring
Social Networks:- Interaction between people
WTD-ICMC-USP
Anything

Most important of all


Content is King


because a network is only useful if it is being used
to carry information that people want
Monitizing content – directly or indirectly - is the
new El Dorado.
A massive paradigm shift in terms of what the
Internet is being used for:

Multimedia, interactive games, real-time
communication in all forms

Low latency financial applications
NOVEMBER 2013
WTD-ICMC-USP
The Key Challenges of Building this
Brave New World


Very different to the current Internet
Support for mobility and location management
is of primary importance

Seamless, uninterrupted communication

Handover must be controlled and managed

Where you are may be used as a hint to where
you may be in the future



Proactive approach
Know where network infrastructure is located
Privacy of Location Information
NOVEMBER 2013
WTD-ICMC-USP
Challenges in Networks

How do we get networks to work together
seamlessly

Heterogeneous Networking

Need a standard for ubiquitous handover between
networks at a low level

Need to allow higher levels (transport
protocol/application) to adapt

Multi-homing: managing all the network interfaces
on a device
NOVEMBER 2013
WTD-ICMC-USP
Challenges in QoS

As traffic is increasing we need to look at some
sort of Quality-of-Service support

Lots of research into Internet QoS models
• IntServ:- failed because it could not scale
• DiffServ:- Works in the core network but not finegrained enough to work on the periphery.
• Need a new approach
NOVEMBER 2013
WTD-ICMC-USP
Challenges in Security

Security must be also a key issue of the Future
Internet

Current Internet is woefully inadequate
•

Multi-level security




Edward Snowden
Authentication, authorization
Denial of Service attacks
Privacy (use of the net not monitored)
Security needs to be built in from the start
NOVEMBER 2013
WTD-ICMC-USP
Challenges of Big Data




New networks are producing data faster than
we can analyse, categorize or process
Storage will also become a big issue
Data Security: who owns my data in the Cloud,
who can access my data, do I have any say
where the data is stored
Results of Big Data
–
Information is now gold
NOVEMBER 2013
WTD-ICMC-USP
Challenges of Service Delivery



Delivering services will also need to change
Services need to be managed in a more
autonomous manner
Spawn new server instances at different
locations or migrate services when required in
response to:
– Geographical load patterns
– User mobility
– Network failure or recovery
NOVEMBER 2013
WTD-ICMC-USP
Challenges of Different Types of
Networking

Opportunistic Networking
–

Taking advantage of social interaction; sensor
networks
Move towards more data-centric paradigms
–
Current Internet still communication-centric
•

Stresses host-to-host communication
Information Centric Networking
–
Getting information – irrespective of location
•
NOVEMBER 2013
Publish and subscribe models
WTD-ICMC-USP
Meeting these challenges

Need an integrated approach
–

Cannot study one challenge in isolation
Failure is not an option
–
Internet is the most successful thing that
humans have made; the wheel is a distant
second
•

Problems are building up and need to be
addressed
Affects all of us
–
No country, institution or company should be in
total control of the Internet
NOVEMBER 2013
WTD-ICMC-USP
How are we going to make this
happen – what is the approach

Revolutionary Approach

Start from scratch
•

Clean Slate Project; Plan 9 test
Evolutionary Approach

Only incremental increases
•

IPv6, INTERNET 2
Challenges mean that an evolutionary jump is
needed
NOVEMBER 2013
WTD-ICMC-USP
How are we going to make this
happen – what is the approach



Try what has worked before
Agree on functionality; but not on
implementation
Need a framework that gives us functionality
but not say how the functionality is
implemented
–
Worked for telephony; 3KHz standard
–
Worked for the Internet: OSI; TCP/IP
NOVEMBER 2013
WTD-ICMC-USP
Why do you need a framework?

To be able to think about the issues coherently

Imposes mental discipline
–
Forces you to always keep the Big Picture in
mind

Separates functionality/policy from mechanism

Frees us to:
•
•
NOVEMBER 2013
Use or enhance existing mechanisms/standards
Only design new mechanisms when needed
WTD-ICMC-USP
PERIPHERAL NETWORK
SECURITY LAYERS
APPLICATION ENVIRONMENTS
QOS LAYER
SAS
QBS
END SYSTEM TRANSPORT
NTS
MOBILITY MANAGEMENT
NAS
HANDOVER MANAGEMENT
NETWORK ABSTRACTION
(MOBILE NODE)
HARDWARE PLATFORM
(MOBILE NODE)
CORE NETWORK
SERVICE PLATFORM
NETWORK QOS LAYER
CORE TRANSPORT
NETWORK MANAGEMENT
CONFIGURATION LAYER
NETWORK ABSTRACTION
(BASE STATION)
HARDWARE PLATFORM
(BASE STATION)
A very brief Introduction to Y-Comm

This is not a talk about Y-Comm
Talks, papers at:
http://www.mdx.ac.uk/research/science_technology
/informatics/projects/ycomm.aspx

In essence Y-Comm is an architecture that is
trying to integrate:
•
•
•
•
NOVEMBER 2013
Communication
Mobility
QoS
Security
WTD-ICMC-USP
Y-Comm: Still a work in progress


It is not the only architecture that is being
studied
–
Ambient networks
–
Mobile Ethernet
Y-Comm is by far:
•
•

The most detailed
The most integrated
Architecture is stable
•
NOVEMBER 2013
Recently tweaked some names of the layers to
make their functionality better understood by the
mobile telcoms community
WTD-ICMC-USP
Why is Y-Comm different


Y-Comm was predicated on two key
assumptions:
Network Evolution
–
The Internet is decomposing into 2
components
•
•

Devices will have multiple Interfaces
•
•

A super-fast core using Optical Switching/MPLS
Wireless Peripheral Networks at the Edge
3/4G, Wi-Fi, WiMax, etc
Called HETNET devices
Both the assumptions turned out to be true
NOVEMBER 2013
WTD-ICMC-USP
Current Internet
Future Internet
BACKBONE
ACCESS NETWORKS
WIRELESS NETWORKS
The Core Framework
SERVICE PLATFORM LAYER
NETWORK QOS LAYER
CORE TRANSPORT SYSTEM
NETWORK MANAGEMENT LAYER
CONFIGURATION LAYER
NETWORK ABSTRACTION LAYER
HARDWARE PLATFORM LAYER
The Peripheral Framework
APPLICATION ENVIRONMENTS LAYER
QOS LAYER
END TRANSPORT SYSTEM
MOBILITY MANAGEMENT LAYER
HANDOVER MANAGEMENT LAYER
NETWORK ABSTRACTION LAYER
HARDWARE PLATFORM LAYER
History of Y-Comm

Pre-Y-Comm (1998-2003)
–
Cambridge Wireless Testbed

2006 – Peripheral Framework announced

2007 – Y-Comm architecture announced

2008 – USP & UFSCar join effort

2010 – Loughborough University joins effort

2012 – Lancaster University joins effort
NOVEMBER 2013
WTD-ICMC-USP
Key People in Y-Comm

Glenford Mapp (Middlesex University)

Jon Crowcroft (University of Cambridge)

Edson Moreira (USP)

Helio Guardia (UFSCar)

Raphael Phan (Loughborough University)

Qiang Ni (Lancaster University)
NOVEMBER 2013
WTD-ICMC-USP
Key PhD students


Fatema Shaikh (Middlesex University 2010)
David Cottingham (University of Cambridge
2010)

Renata Porto Vanni (USP 2010)

Mahdi Aiash (Middlesex University 2012)

Rigolin Lopes (USP 2012)

Mario Augusto (USP 2012)

Fragkiskos Sardis (Middlesex University)

Ann Samuels (Middlesex University)
NOVEMBER 2013
WTD-ICMC-USP
Key Middlesex MSc Students

Diti Dave (2010)

Naveen Chinnam (2011)

Ali Mofidizati (2012)

Rajesh Lakkineni (2012)

Brian Ondiege (2012)

Eghe Akenuwa (2013)

Eric Ghokeng (2013)
NOVEMBER 2013
WTD-ICMC-USP
What are the major contributions of
Y-Comm so far

Handover
–

Security
–

Integrated Security; Targeted Security Models;
Ontologies for Communication Architectures
–

Handover Classification; Proactive vertical
handover, Calculations for NDT and TBVH
Y-Comm Ontology, MyHand
Quality-of-Service
–
New QoS Framework
NOVEMBER 2013
WTD-ICMC-USP
Things being worked on – NOT part
of this talk

An implementation of IEEE 802.21
•

Game Theory in Communication Systems
•

To see if game-theory can lead to optimum
resource allocation (Lancaster University)
A new transport protocol for LANs
•

To provide seamless handover (UFSCar)
To optimize server speeds in LANs and Clouds
(Middlesex University)
A Hybrid Internet QoS model
•
NOVEMBER 2013
Combining IntServ and DiffServ (Middlesex
University
WTD-ICMC-USP
End of Y-Comm Part

Y-Comm has been a success because it has
provided a framework to allow us to begin to
exploring how to support the 4As
–

Just a start; still a very long way to go before
we get to this new world
Y-Comm has moved from the
design/architectural phase to the
implementation phase
–
Exploring using Software Defined Networking
(SDN) as a way of building a full Y-Comm
prototype
NOVEMBER 2013
WTD-ICMC-USP
Questions on Y-Comm
NOVEMBER 2013
WTD-ICMC-USP
Going Deep


In this section we want to look at how a section
of the work in Y-Comm is impacting 3 key
areas:
–
Resource Management in Core Networks
–
Mobile Services
–
VANETs
Need to understand proactive handover in YComm
NOVEMBER 2013
WTD-ICMC-USP
Basic Handover Terms



Hard vs Soft Handover
–
Hard:- break before make
–
Soft:- make before break
Network vs Client Handovers
–
Network-based
–
Client-based (Apple's Patent)
Upward vs Downward
–
Upward – smaller to bigger coverage
–
Downward – bigger to smaller coverage
NOVEMBER 2013
WTD-ICMC-USP
Handover Classification
HANDOVER
ALTERNATIVE
IMPERATIVE
SERVICES
NETPREF
REACTIVE
UNANTICIPATED
PROACTIVE
ANTICIPATED
KNOWLEDGE-BASED
MODEL-BASED
USERPREF
CONTEXT
Knowledge-Based Proactive Handover
(Cambridge)
Model Based Proactive Handover




The work of Fatema Shaikh
Define a circular area of coverage called the
Handover radius
Define a smaller radius called the Exit Radius
at which handover must start in order for the
handover to be completed at the Handover
Radius
The time the mobile node has before it hits the
Exit Radius is called Time Before Vertical
Handover or TBVH
NOVEMBER 2013
WTD-ICMC-USP
Model-Based Handover
Handover
threshold circle
Exit threshold
circle
Threshold Circle
coverage
Real coverage
Exit coverage
NOVEMBER 2013
WTD-ICMC-USP
Predictive Mathematical Model for TBVH
(Simple Case)
Movement of MS under BBS coverage (upward vertical handoff)
•
Introduction of additional functionality
to Base Station at
network boundary (BBS).
•
Distance between MS and BBS
derived from location
co-ordinates or
•
Estimated TBVH
Simulation and Results
TBVH simulation in OPNET Modeler:
Why is TBVH important

If the Mobility Management Layer can
calculate TBVH, it can signal to the higher
layers that a handover will occur after a certain
time so these layers can take action.
–
Minimize the effects of handover delay and
packet loss by buffering and using fast
retransmission techniques
–
It makes proactive handovers more seamless
compared to reactive handovers
•
NOVEMBER 2013
Can Fatema Shaikh's work be extended to any
arbitrary situation?
WTD-ICMC-USP
Combining Transport and Communications to
determine the optimum handover
NET A
A
NET B
S
B
NET C
C
T
Analysis shows that it is possible to calculate
these key points with some degree of accuracy
C1
E1
Y2
Z1
A
S
Y1
H1
Y3
C2
B
E2
C
E3
H3
H2
T
Z2
Z3
Results
What does it all mean?

If the mobile node knows: its location, direction
and velocity
–

The location of the networking infrastructure
–


Via GPS or accelerometers
Type of access network, the position of the
access points
Good estimation of the Handover Radius
Then we can calculate the optimal times to
handover over a large region (a few miles)
NOVEMBER 2013
WTD-ICMC-USP
REQ (Time , TBVH, NDT)
A
A
WIRELESS
NETWORK
B
REQ (Time , TBVH, NDT)
B
Rethink Allocation Strategy




MNA needs channel at (Time + TBVH) A
MNA releases channel at (Time + TBVH +
NDT)A
MNB needs channel at (Time + TBVH)B
MNB releases channel at (Time + TBVH +
NDT)B
NOVEMBER 2013
WTD-ICMC-USP
There are 3 possible outcomes

No contention:
•
•

(Time + TBVH)A < (Time + TBVH)B
(Time + TBVH + NDT)A < (Time + TBVH)B
Contention: Two Types: Partial and Total
• (Time + TBVH)A < (Time + TBVH)B
• (Time + TBVH + NDT)A > (Time + TBVH)B
• Partial Contention:
(Time + TBVH + NDT)A < (Time + TBVH + NDT)B
•
Total Contention:
(Time + TBVH + NDT)A >= (Time + TBVH + NDT)B
NOVEMBER 2013
WTD-ICMC-USP
Request Summary



Requests granted as requested:
–
Channel granted at (Time + TVBH)A
–
Channel released at (Time + TBVH + NDT)A
–
Same with B
Requests granted but modified for B
–
Channel granted at (Time + TBVH + NDT)A
–
Channel released at (Time + TBVH + NDT)B
Request for B not granted:
–
Force B to handover to another network
NOVEMBER 2013
WTD-ICMC-USP
Further Results




Simulation results show that there is a clear
benefit to using this approach
Nodes that can use the channel are not forced
to wait behind nodes that cannot use the
channel
Nodes that cannot use the channel quickly
handover to other networks so we avoid
unnecessary handover attempts
Good result for operators
NOVEMBER 2013
WTD-ICMC-USP
Mobile Services



PhD research by Fragkiskos Sardis
General idea: as the users of a service move
around the latency between the user and the
service could increase such that the user's
QoE could be affected
A way around this is to be able to migrate or
replicate the service at a location closer to the
mobile user to reduce latency
–
Cloud Services now make this possible
NOVEMBER 2013
WTD-ICMC-USP
Key Issues

What is the condition that indicates that
movement of the server should be considered
–



Latency/Bandwidth Threshold
What is the algorithm which decides to which
Cloud should the service be migrated or
replicated
There is a cost for moving the service which
cannot be ignored
We need a service delivery framework
NOVEMBER 2013
WTD-ICMC-USP
The Service Delivery Framework
SERVICE MANAGEMENT LAYER
SERVICE SUBSCRIPTION LAYER
SERVICE DELIVERY LAYER
SERVICE MIGRATION LAYER
SERVICE CONNECTION LAYER
NETWORK ABSTRACTION LAYER
Key Observation

Whether I move the service to a Cloud on the
network to which the user is currently
connected depends on how long the user is
expected to be in that network
–

Network Dwell Time or NDT
So the win is:
–
The bytes saved running the service locally –
the bytes needed to migrate the service
–
First parameter is dependent on NDT
NOVEMBER 2013
WTD-ICMC-USP
How could we test this



We set up a gaming scenario between a
mobile device and two Clouds.
We begin playing the game on one Cloud and
then signal to the system to migrate the game
to the second Cloud.
When we migrate to the new network, we
measure the bytes saved vs the amount of
time I am in the network of the second Cloud
NOVEMBER 2013
WTD-ICMC-USP
Results
2000
1500
1000
500
0
1080
1140
1200
1260
1320
-5 0 0
-1 0 0 0
-1 5 0 0
-2 0 0 0
-2 5 0 0
-3 0 0 0
NOVEMBER 2013
WTD-ICMC-USP
T ra ffic s a vin g s
Further Work


NDT is crucial for good migration of services
We are developing a Markov Chain model for
service migration

Looking at caching content as well

We now move on to the final part of this talk
–
Seamless communication in VANETs
NOVEMBER 2013
WTD-ICMC-USP
VANETs

VANETs are important because they are the
key components for building an Intelligent
Transportation Infrastructure (ITS)
–

ITS involves the integration of the
Communication and Transport Infrastructure
Improve safety on the road
–
Crash Avoidance, Accident Notification

Infotainment in vehicles

ITS is part of Smart Cities research
NOVEMBER 2013
WTD-ICMC-USP
VANET Infrastructure


VANET Hardware
–
Roadside Units (RSUs)
–
Onboard Units (OBUs)
V2I communication
–

V2V communication
–

RSU ↔ OBU (Middlesex University)
OBU ↔ OBU (USP)
Beaconing V2I or V2V
NOVEMBER 2013
WTD-ICMC-USP
Seamless V2I



The work of PhD student Arindam Ghosh
How do you guarantee seamless
communication between the car and the
infrastructure
Seamless handover situation but slightly
different
–
Velocities may be high
–
Beaconing effect
•

Frequency and size of beacon
Interested in proactive handover
NOVEMBER 2013
WTD-ICMC-USP
Our Approach is to look at 3 phases
of Communication


Data Exchange Phase
–
Not concerned about handover
–
Can only see 1 RSU
Doing calculation for handover
–


Determining Exit Time to start handover
Doing the Handover
All must be done before I leave the first
network, i.e., before NDT expires
NOVEMBER 2013
WTD-ICMC-USP
Scenario
NOVEMBER 2013
WTD-ICMC-USP
Use Y-Comm Approach to Calculate
NDT

A good approximate for NDT is
–
NDT ~ 2R/v
•
•


R = Radius of coverage of RSU
V = velocity of the vehicle.
This is an ideal NDT because it is only based
on coverage, it assumes no interference
between RSU and RSU or OBU and OBU
Measure NDT using simulation and find out
how close we get to the ideal NDT and how
this result is affected by other factors
NOVEMBER 2013
WTD-ICMC-USP
Simulation Scenario
NOVEMBER 2013
WTD-ICMC-USP
Results show that:




The higher the beaconing frequency the
greater the NDT from 1Hz ->10 Hz
The size of the beacon also affects the
measured NDT
Velocity also affects NDT severely. The greater
the velocity the less NDT is available.
Also measured the effect of velocity on datatransfer rates (non-linear)
NOVEMBER 2013
WTD-ICMC-USP
Data Exchange Rates at Different
Speeds
NOVEMBER 2013
WTD-ICMC-USP
Summary

VANET is a new area for us

Using NDT is a new appoach

Interesting results

Hoping to develop an analytical model

Need to consider other factors
– Traffic density, interference, etc.
NOVEMBER 2013
WTD-ICMC-USP
Joining the Revolution

Join a work that has just started
• Hybrid QoS Internet Model
• IEEE 802.21 implementation
• Proactive Channel Allocation
• VANET

Or explore a total new area in Y-Comm
• Programmable Networking
• SDN and Open Flow
• ICN
NOVEMBER 2013
WTD-ICMC-USP
THANKS FOR LISTENING
ANY QUESTIONS?
NOVEMBER 2013
WTD-ICMC-USP