Transcript Introduction to Course

Designing Future Networking
Systems
Shaping Future Telecom Operators.
a project course by the members of detusche telekom laboratories
Designing Future Networking Systems.
1
Designing Future Networking Systems.
2
Designing Future Networking Systems.
A Project Course by the Members of Deutsche Telekom
Laboratories.
•Clean Slate Internet Design
•What are the current problems in internet architecture?

What are the proposed solutions?
•What is the vision for future networking systems?

Much more than Internet 2.0!

Ubiquidous high-speed wireless access

Secure host identification / Secure anonymization

Robust routing and transport delivery

Session management that works!
•Help us design the future
•Topics

Application Layer
– Service Placement

Transport Layer
– Host Identification Protocol (HIP)
– Adaptive Queue Management (AQM)
– Cross-layer TCP for wireless links
– Heterogenous Access Networks

Routing Layer
– Probabilistic Routing
– Open Routers
3
Designing Future Networking Systems.
Course Administration.
•Course website

https://www.dai-labor.de/index.php?id=580
•Course administrator

Martin Roth

[email protected]
•12 weeks
• First 4 weeks introduction of concepts and technologies
• Next 7 weeks, project specific lectures, milestone meetings
• Last 1 week, final project presentation, demonstrations
• Final reports thereafter
•One report per project

about 10 pages

Technology review, implementation details, experience

in English!

Talk to us...
– Templates available
•Course Evaluation

70% Project

20% Presentations

10% Report
4
Application Layer.
Service deployment platforms
by Evangelos Kotsovinos
•Service deployment platforms allow users to
obtain resources on machines they do not
directly own or control (e.g. Grids, PlanetLab,
Utility Computing)
•XenoServers are servers able to safely host and
execute services provided by third parties
(service providers) in exchange for money
•Services can be deployed on demand, acquiring
computing resources dynamically
•Services may migrate on demand, based on
changes e.g. in network conditions, geography of
client demand, or pricing
•IBM, Sun, HP, Deutsche Telekom, Amazon.com
sell or aim to sell similar facility
•Objectives
 Learning more about deployment technologies
 Comparing the different deployment approaches
 Discussing application scenarios for which each type of approach is more appropriate
6
Network Design
xCAT (Cross Capacity Analysis Tool).
How can we prepare for 4G systems?



Objective
– To study the challenges in the network planning of future communication
systems (i.e. 4G networks)
Cross system engineering
– The set of rules that define the cooperation and competition among the
different access networks within a 4G system
Tasks: To extend the current simulation tool and develop different optimisation
algorithms
8
xCAT (Cross Capacity Analysis Tool).
How can we prepare for 4G systems?

Relevance
– Future integrated operators will require rules to coordinate interactions
among the different networks
– These interaction rules are strongly related to the business models
– These rules are an important element for self-optimised systems
– Self-optimised systems reduce network management and deployment
costs for Deutsche Telekom
9
Transport Layer.
TCP-FAT (Fast Adaptation Time).
Impact of Mobility on the transport layer.


Objectives
– To analyse the impact of vertical handovers on TCP connections
– To characterise the effects (adaptation delay component, Ta)
– To design techniques that reduce these effects
– To evaluate our proposal
Tasks
– Experimental setup
– Collect traces
– Adapting existing scripts
– Working on a new definition for Ta
– Analysing the traces
11
TCP-FAT (Fast Adaptation Time).
Impact of Mobility on the transport layer.

Relevance
– TCP traffic represents 40--60% of the traffic in the Internet
– We need to support real-time services on the move
– We need to minimise delays (latency) everywhere
– A cross-layer solution that tackles mobility at different layers is
needed to enable seamless networking
– Reducing handover latency (network layer) is not enough
– Reducing adaptation delay in vertical environments is
fundamental for future mobile scenarios
12
Efficient Scheduling across air-interfaces

Multiple Interfaces: allows advantages of technology
diversity.
– WLAN : high bandwidth (but low mobility)
– Cellular: intermediate mobility support (but
low bandwidth)

Wireless Access
via WiFi, cellular
satellite solutions
Simple bandwidth sharing can lead to low throughput
and waste network resources
13
An Integrated Approach for
TCP Throughput Optimization
Model TCP’s congestion control dynamics and overlay an optimization framework
Using Dynamic Programming (DP) principles, evaluate optimal throughput for a bulk transfer TCP flow
Investigate Link/PHY layer adaptation for throughput optimization, via

Power control: crucial and usually indispensable for wireless networks

Link Adaptation: Proven merits (e.g. WLANs, 802.11n, 802.16 proposals, EGPRS)
14
Resiliency Measures for a Tree-based Overlay Structure
15
SOUND-NET.
Unveiling User’s Perception of Future Communications.

Objectives
– To asses user experience in future 4G networks
– To design mobility tests (targeting VoIP)
– To evaluate the scenarios
– To extend the e-model for

Handover, technology switching, NB  WB, etc
– Extract the appropriate planning information
16
SOUND-NET.
Unveiling User’s Perception of Future Communications.

Relevance
– Seamless mobility does not mean zero-disruptions
– Always-best-connected needs to be evaluated
– We need to know how the user may perceive seamless services
– We need to know users’ perspective in order to design future services
and supporting resources
17
Host Identification Protocol (HIP).
Supporting secure mobility
Background
Host Identification Protocol (http://www.ietf.org/html.charters/hipcharter.html), is considered to be the next big thing in the Mobile
Internet landscape as it combines mobility management elegantly with
security, in particular, authentication and encryption. It provides
methods of separating the end-point identifier and locator roles of IP
addresses, as well as introduces a new name space, Host Identity,
based on the public keys system.
Project outline
In this project, the aim is to firstly survey existing publicly available
experimental HIP implementations and secondly gain unique hands-on
experiences in setting up HIP in heterogeneous networks
environments.
Tasks
You will be given a unique opportunity to setup a test-bed environment
capable of switching an incoming music streams between any IP
enabled devices using different access technologies.
18
Routing Layer.
Wireless Mesh Networks
Wireless, infrastructure-based mesh networks promise
 Fast deployment
 Cheap deployment (compared to fiber)
 High data rates (scalable)
Various use scenarios
 Developing countries (China, India): an
infrastructure
 Well-connected countries (Korea): ubiquitous
access
Possible deployments in Berlin/ Germany
 Biergarten, Coffee shops, shopping areas
 Parks (Tierpark, lakes around Berlin)
 Neighborhoods (garden, common grounds, EastBerlin)
 Ski stations
WiFi mesh
Internet
20
Open Source in the Context of Routing Platforms
Why is it of interest?

Research community:
– Allow real-world experimentation and evaluation of network protocols
– Enable the development of router applications
– Facilitate novel designs in network element and protocol stack architecture
– Act as a way to avoid Internet ossification – provide a path for adoption and
deployments

Operator community:
– Decrease TCO for network equipment
– Enhance interoperability among network elements from different equipment vendors
– Avoid network equipment vendor lock-in
– Decrease time-to-market for new network services, bug fixes, etc.
The multiple layers of “openness”

Open platforms: provide clear and well-defined interfaces (programming abstractions) for
developing and integrating new protocols and system components (e.g. management
interfaces, schedulers, forwarding paths, etc.)

Open protocol stacks: expose the internals of network protocols allowing the development
of new features, extensions and modifications
– Open device drivers: usually enable modifications and tweaking with the link-layer
and medium access layer of the network access technology

Open hardware: make available the details of the reference design and the hardware
abstraction layer, allowing arbitrary ways of accessing the underlying hardware
21
Node Hardware
RouterBoard from Mikrotik.com
 Fast main board in small-form-factor
– MIPS 32 4Kc based 266MHz
– 64/128Mb RAM, 1Mbit for bootloader
– 3-8 Ethernet 10/100Mbits cards,
PoE
– One serial RS232c port
 For Magnets II
– CompactFlash cards (2Gb max)
for trace collection
– MiniPCI bus: 2-6 slots for MiniPCI
devices:

Up to 6 cards: WiFi, GSM,
UMTS, Bluetooth, Zigbee,
later WiMAX
 Open-source Linux platform
– Linux 2.4, patch for bootloader
included
22
Probabilistic Routing.
by Martin Roth
•Traditional Routing is Deterministic
 Link-State (Bellman-Ford)
 Distance Vector (Dijkstra)
•Paths are Brittle
 Explicit Multipath Routing is
necessary for Robustness
•Network (Re)configuration is Expensive
 Lots of Control Traffic Overhead
is required
•Isn‘t There a Better Way?
•Probabilistic Routing!
 Generate a probability distribution
over every path in the network
 Route according to path utility
23
Swarm Intelligence.
Biologically Inspired.
24
Questions?
Fragen?
Martin Roth [email protected]
Pablo Vidales [email protected]
Jatinder Pal Singh [email protected]
Students
Sebastian [email protected]
Rafael [email protected]
Niklas [email protected]
25
Host Identity Protocol (HIP)
Robert Hsieh 26th April 2006
Background of the origin of HIP
IP Address serve the duel role of being
 End Point Identifiers
– Names of network interfaces on hosts
 Locators
– Names of naming topological locations
This duality makes thing very hard!! IRTF Name Space
Research Group debates for years without reaching
consensus
27
HIP in a Nutshell
Integrates security, mobility and multi-homing
 Opportunistic host-to-host IPSec ESP
 End-host mobility across IPv4 and IPv6
 End-host multi-address multi-homing across IPv4 and
IPv6
 IPv4 and IPv6 interoperability for apps
A new layer between IP and Transport layers
 Introduces Cryptographic Host Identifiers
28
The HIP Project
Survey existing publicly available HIP implementations
Install and test the various HIP implementations for
comparison
Select one HIP implementation and design various
demonstration scenarios to showcase its capability and
possible drawbacks
29