Wireless software defined networks: 5G or beyond 4G… Dr. Yoram haddad Jerusalem College of Technology- Lev Academic center [email protected] cc.jct.ac.il/~haddad ©Yoram Haddad-2013-2014 DST - CSIR - South Africa, March 2014
Wireless software defined networks: 5G or beyond 4G…
Dr. Yoram haddad
Jerusalem College of Technology-
Lev Academic center [email protected]
cc.jct.ac.il/~haddad
©Yoram Haddad-2013-2014
DST - CSIR - South Africa, March 2014
80 B€ for 2014-2020
Expected impact in “ICT 14 – 2014: Advanced 5G Network Infrastructure for the Future Internet”
1000 times higher mobile data volume per geographical area
Network architecture, protocols and radio technologies capable of at least a ten times increase in frequency reuse
10 times to 100 times higher number of connected devices
10 times to 100 times higher typical user data rate
10 times lower energy consumption for low power Machine type communication
5 times reduced End-to-End latency
Ubiquitous 5G access including in low density areas
©Yoram Haddad-2013-2014
Interesting facts
Wireless Spectrum overcrowded but...
Only 2% of wireless spectrum between 30 Mhz and 3Ghz effectively used (measured in 2009 in dense urban area near Washington DC, USA)
In UK more people connected to broadband than Water and Average utilization of internet link is….<1% but…
Electronics and Software industries spend billion of dollars for verification and what about telecom?
Surrounded by huge number of wireless AP and cellular BS from different operators and still not enough big MOBILE data rates
©Yoram Haddad-2013-2014
Estimation: Cellular traffic will double approx. every 9 months
Qualcomm CEO, Paul Jacobs in 2008: “improvement of wireless links that enhance user throughputs is reaching its limit… compare Wireless link…HSPA+ with that of LTE, their difference in frequency utilization efficiency is only about 10%. It is small
enough to be offset by noise…. Solution: densely deploying base stations to shorten the distance between base stations and mobile terminals
4G LTE PHY spectral efficiency within 20% of Shannon Capacity limit (source: ONRC research)
Base Station density will have to increase 20x
Increase wireless link efficiency ?! 5G ?
©Yoram Haddad-2013-2014
Evolution of underlying core protocols (TCP,IP) and Algorithms (congestion,…) limited
Vs. ↑ number of internet users
Vs. smartphone openness revolution
Vs. Better access link e.g. FTTH
Why ?
Gradual deployment
Test on real testbeds not simulation
This is critical to avoid network ossification
Motivation (1): Innovation
©Yoram Haddad-2013-2014
Internet is a best effort network
Providing Quality of Service remains a challenge
Users connected everywhere, everytime but no seamless service continuity e.g. movie begins on train to finishes at TV home
Why?
Seamless handover between operator and networks requires central management entity that
Have visibility over all relevant components
Identify, adapt and reroute flow between end points
Motivation(2): Performance
©Yoram Haddad-2013-2014
Checking correct behavior of network is difficult
But…operators must check continuously loop free routing, reachability, …
Up to now, black art -> “by hand”
Fallback mechanisms to solve problems
Why?
No formal method and verification tools
Information must be gathered to a central entity
Motivation(3): Verification
©Yoram Haddad-2013-2014
Vertically integrated
Closed, proprietary
Slow innovation
Small industry
Specialized
Operating
System
Specialized
Hardware
App App App App App App App App App App App
Specialized
Applications
Horizontal
Open interfaces
Rapid innovation
Huge industry
Microprocessor
Open Interface
Linux Mac
OS
Windows
(OS) or or
Open Interface
[Source: Prof Nick Mckeown,
ONS 2012]
Vertically integrated
Closed, proprietary
Slow innovation
App App App App App App App App App App App
Horizontal
Open interfaces
Rapid innovation
Control
Plane
Control
Plane
Control
Plane or or
Open Interface
Specialized
Control
Plane
Specialized
Hardware
Specialized
Features
Merchant
Switching Chips
Open Interface
[Source: Prof Nick Mckeown,
ONS 2012]
In 2007 seminal research papers propose decoupling forwarding plane (physical infrastructure) and control plane
In the past (present) …
Routers no longer decide but only classify entering flow
Action decided by a remote central entity : “controller”
Flow: set of packets with same characteristics (among 12-tuple header’s field)
Communication between devices and controller via OpenFlow protocol
Software Defined Network
©Yoram Haddad-2013-2014
Traditional Control Mechanisms
Distributed algorithm running between neighbors complicated task-specific distributed algorithms
Source: Prof. Scott Shenker
Control Program
Software Defined Network (SDN)
Network OS
Global Network View
e.g. routing, access control
Source: Prof. Scott Shenker
Define and Upload rule/action pairs into interconnection equipment
Provides execution environment for programmatic control of the network Network Operating System (NOS)
Open source controller: NOX in C++, POX in Python
Devices (switches, routers,..) send their state (no of packets of given flow that entered etc..) to controller
Update policies dynamically based on this information
Controller and Network OS (NOS)
©Yoram Haddad-2013-2014
Slicing
Flow IDs for experimental traffic without interference to production network
Slicing done by FlowVisor
Illusion : each controller to dedicated network
Long Term vision: decoupling between “physical infrastructure” and “service provider”
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Network Planning
• Goal: allocate channels so that neighboring APs use different channels
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Dynamic change to the best channel
Close APs should use different channels so that they don’t interfere Planning, see above Close access points often belong to different
operators anyway...
An AP could sense the surrounding channels
occupancy and take a greedy local decision Switch to the less busy channel
Issue #1: clients have to switch too It could break network connections => don’t do this
often
Issue #2: what if other APs apply the same policy? May be useless, or cause oscillations
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
How can network virtualization help?
• A good percentage of Wi-Fi access points are set-top-boxes • Owned by the ISP
• In a single operator scenario:
• All access points can report measurements to an area controller (building, ...) - e.g. use IEEE 802.11v extension
• Can also report usage statistics (users signal attenuation distribution, ...) • The controller runs a planning algorithm (under constraint) • The controller uploads a policy (channel, TX power, ...) to the access points
• In a multi-operator scenario • Same strategy, except that controllers have to communicate together
• Bonus: • Inter-operator QoS management becomes possible
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Wireless SDN proof of concept: redundant transmission
• WLAN has a limited throughput and are sensitive to congestion • Global performance is disappointing when the user is used to wired LAN • Multimedia content is tailored for Ethernet LANs
• Solution at the signal level: MIMO • Solution at the packet level: duplicate the packet and
send it through different paths • Requires a multi-interface terminal
• Can work across technologies (e.g. Wi-Fi and
LTE smartphone)
• Scenario implemented as an OpenRoads demo
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Cross-layer redundant transmission for multimedia applications
Multimedia flows are sometimes composed of different types of frames
Example: MPEG flow I-Frames: full images P-Frames: “predicted” images (contains only
differences from the previous image) B-Frames: “bi-directionnal predicted” images
(differences between the previous & the next images)
Loss or delay does not have the same effect on all types of frames
Where SDN can help Transmit I-Frames on the best channel,
P and B frames on the other Adapt video quality to network conditions
(filter detail frames, ...)
©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Handover management
• Horizontal handover (from AP to AP) • Channel measurements ease mobility prediction
• SDN enables dynamic redefinition of forwarding rules to duplicate traffic over two APs
- Need mechanism to filter duplicate packets at the reception
• Vertical handover (between technologies) • Similar technique as horizontal handover when controllers
collaborate
• Mobility prediction requires the device to decide whether to switch networks
• Scenarios are already demonstrated by OpenRoads ©Yoram Haddad-2013-2014
FP7-Crossfire Workshop, July 2013
Inter-ISPs roaming and handover
• Scenario: two (or more)
partner ISPs offer Wi-Fi seamless roaming
©Yoram Haddad-2013-2014
Buzzwords
CloudRAN (RAN = Radio Access Network)
RANaaS
CloudNetworking
Cellular IaaS
Source: Blueprint for Introducing Innovation into Wireless Mobile Networks by K.K. Yap et al.
©Yoram Haddad-2013-2014
[source: Google ONS
presentation,April 12]
Not Sci-Fi…
Run at close to 100% utilization instead of 30-40 % usually [Amin Vahdat, Google Principal Engineer, June 2012 ]
Academy Stanford & Berkeley : Open Networking Research Center
Princeton
…
Industry NEC
Google (Openflow deployed in its dedicated backbone)
Deutsche Telekom (with TU berlin)
Nicira by Vmware
…
Other ONF
Players…
©Yoram Haddad-2013-2014
SDN is unavoidable
But at a small scale level:
mmWave
Massive MIMO
WiFi direct or device to device or Ad-hoc
Unlimited EIRP and Game theory for power control
4th channel for wifi
And…
So 5G…
©Yoram Haddad-2013-2014
Encryption of physical signal
Ad-hoc and mesh over SDN
SDN controller performance
Cellular SDN backhaul
Video splitting over multiple radio for Medical
Efficient deployment of cellular and wifi
Turn off “sleeping” AP and/or BS
Power control through cluster
..
Other projects
©Yoram Haddad-2013-2014