A survey on Wireless Mesh

NetworksIF Akyildiz, X Wang - Communications Magazine, IEEE, 2005

Youngbin Imybim@mmlab.snu.ac.kr

2007.10.15.

Contents• Introduction to WMNs• Network architecture• Critical design factors• Advances and research challenges▫ Physical layer▫ MAC layer▫ Routing layer▫ Transport layer▫ Cross-layer design

• WMN examples

Introduction to WMNs (1/4)• Mesh networking defined in “Wikipedia”▫ A way to route data, voice and instruction b/w nodes▫ Allows for continuous connections and reconfigurationaround broken or blocked paths by “hopping” from node to node

▫ Difference from other networks� The component parts can all connect to each other via multiple hops

� Generally not mobile▫ one type of ad hoc network

Introduction to WMNs (2/4)• Two types of nodes▫ Mesh routers

� Minimal mobility� Form the mesh backbone � compared with conventional wireless routers

� Additional routing functions for mesh networking� Usually multiple wireless interfaces, but similar hardware

▫ Mesh clients� Simpler hardware and software

Introduction to WMNs (3/4)• Integration of WMNs with other networks

▫ Through the gateway/bridge functionalities of mesh routers

▫ Difference from typical ad hoc networks

• Diversify the capabilities of ad-hoc networks

• Application scenarios▫ Broadband home networking

▫ Community networking

▫ Building automation

▫ High-speed metropolitan area networks

▫ Enterprise networking

Introduction to WMNs (4/4)• Some products, some testbeds• But, need considerable research effort▫ Ex) scalability

▫ Researchers revisit the protocol design of existing wireless networks in WMNs’ view� 802.11, ad hoc networks, wireless sensor networks

▫ Industrial standards groups are working on new spec for WMNs� IEEE 802.11, IEEE 802.15, IEEE 802.16

Network architecture• Infrastructure/Backbone WMNs

▫ Provides a backbone for conventional clients ▫ Enable integration of WMNs with existing wireless networks

• Client WMNs▫ Provides peer-to-peer networks among clients▫ No mesh router, same as a conventional ad hoc network

• Hybrid WMNs▫ Combination of both▫ clients access through mesh routers or other mesh clients directly

critical design factors (1/2)• Radio techniques

▫ directional and smart antennas

▫ MIMO

▫ Multi-radio/multi-channel systems

• Scalability - all layers

• Mesh connectivity

• Broadband and QoS

▫ Most applications are broadband services

▫ Need more performance metrics

� besides delay, fairness

� delay jitter, aggregate & per-node throughput, packet loss ratio

critical design factors (2/2)• Security▫ Many schemes of wireless LANs or ad hoc are not fully applicable for WMNs

• Ease of use▫ Autonomous network▫ Network management tools

• Compatibility and Inter-operability▫ Backward compatibility with conventional client nodes▫ Integrating heterogeneous wireless networks

Physical layer• adaptive error resilience through link adaptation – modulation, coding rates

• For high-speed transmission

▫ Orthogonal frequency multiple access (OFDM)

▫ Ultra-wide band (UWB)

• For capacity increase, impairment mitigation

▫ Antenna diversity, Smart antenna, MIMO

• For better spectrum utilization and viable frequency planning

▫ Frequency-agile radios, Cognitive radios

• A software radio platform is promising technique for physical layer

• Open research issues

▫ Further improve the tx rate

▫ Interactive working with higher level protocols� Cross-layer design, low-cost software radio

MAC layer (1/2)• Single channel MAC approaches

▫ Modifying existing MAC protocols

� ex) Adjust parameters of CSMA/CA, but low throughput

▫ Cross-layer design

� Directional antenna-based MACs

� MACs with power control

▫ Innovative MAC protocols

� TDMA or CDMA MAC protocols

� But, complexity, cost, compatibility with existing MACs

MAC layer (2/2)• Multi-channel MAC approaches

▫ Multi-channel single-transceiver MAC� For cost, compatibility

▫ Multi-channel multi-transceiver MAC� Only one MAC

▫ Multi-Radio MAC� Each radio has its own MAC, PHY� Multi-radio unification protocol(MUP) – coordinate communications in all channels

• Open research issues▫ Scalable MAC - Need new distributed & collaborative schemes▫ MAC/Physical cross-layer design

� Novel MAC protocols for advanced PHY techniques – MIMO, CRs▫ Network integration in the MAC layer

� Advanced bridging functions for different wireless radios to work seamlessly

Routing layer (1/2)• Routing protocols with various metrics

▫ In [6], link quality source routing(LQSR) � selects a routing path according to 4 metrics

� ETX, per-hop RTT, per-hop packet pair, hop count � ETX : best for stationary nodes� Hop count : best for mobile nodes

• Multi-radio routing▫ A multi-radio LQSR(MR-LQSR) is proposed in [7]

� Weighted cumulative expected transmission time (WCETT) –Consider link quality metric & the minimum hop count

• Multi-path routing▫ Select multiple paths b/w source and dest▫ Use another path when broken▫ better load balancing & high fault tolerance▫ But, depends on the availability of node-disjoint routes▫ complexity

Routing layer (2/2)• Hierarchical routing▫ Form clusters. Select cluster head which communicate with other clusters

▫ Better performance in high node density ▫ complexity of maintaining the hierarchy

• Geographical routing▫ Use the position of nodes in the vicinity & the destination

▫ Unlike early single-path greedy routing, planar-graph-based algorithm[9] guarantee delivery, but higher communication overhead

Transport layer (1/2)• No proposed transport protocol for WMNs• But, many protocols for ad hoc networks • Reliable data transport

▫ TCP variants - Tackle these problems� Non-congestion packet loss� Unknown link failure� Network asymmetry

� Different BW, loss rate, latency between forward & backward paths� Impact ACKs

� Large RTT variations� The change of routing path may be frequent� The normal operation of TCP relies on a smooth measurement of RTT

▫ New transport protocols� ATP(ad hoc transport protocol)

� Rate-based transmission, Quick-start for initial rate estimation� Delay-based congestion detection, No retransmission timeout

� Better performance than TCP variants� Not compatible with TCP

Transport layer (2/2)• Real-time delivery▫ A Rate control protocol(RCP) is needed to work with UDP

▫ An adaptive detection rate control(ADTFRC) scheme� Multi-metric joint detection approach for TCP-friendly rate control schemes

• Open research issues▫ Cross-layer solution to network asymmetry

� Should avoid asymmetry b/w data and ACKs at routing layer

� MAC & error control need to treat TCP data & ACKsdifferently to reduce asymmetry

▫ Adaptive TCP � Which can be used in heterogeneous networks

Cross-layer design• Protocols should interactively work together with especially with the PHY

▫ Channels are variant in terms of capacity, bit error rate, etc

▫ Dynamic network topology impact multiple layers

• Two ways

▫ By considering parameters in other layers

� Ex) the packet loss rate in MAC is reported to transport layer – differentiate congestion from loss

� Keep the transparency

▫ By merging several protocols

� Ex) MAC & routing

� Better performance

• Issues

▫ Loss of layer abstraction

▫ Incompatibility with existing protocols

▫ Unforeseen impact

▫ Difficulty in maintenance

RoofNet• An experimental 802.11b/g mesh network in development at MIT CSAIL (computer science and artificial intelligence laboratory)

• open source software▫ Join the existing RoofNet

▫ Start a new community network

• Rice University's Networks Group has partnered with Technology For All (TFA) to deploy a multi-tier wireless network in an under-resourced community in Houston's East End.

▫ Stared in 2003

▫ 3,000 users in summer of 2007

▫ Fully programmable nodes

▫ a first-of-its-kind research testbed for large-scale urban wireless networks

TFA(Technology For All)

Meraki• A company in Mountain View, California

• Mission : bring affordable Internet access to the next billion people.

• 25 countries

• How it works

▫ Some offer APs

▫ The connections are extended by repeaters to neighbors

• Portland, Oregon example

▫ A hundred routers, $4,999

▫ 1000 people, pay $1 per month

▫ By 5 DSL connections

▫ Same broadband quality with the normal connection

XO-1• known as the $100 Laptop or Children's Machine

• One Laptop per Child (OLPC) social welfare organization

• an inexpensive laptop computer intended

▫ to be distributed to children in developing countries

▫ to provide them with access to knowledge.

• Currently set to start at $188

• Goal : $100 in 2008.

• IEEE 802.11b, set maximum to 2 Mb/s

• will use IEEE 802.11s to form the wireless mesh network.

• How to access to the Internet

▫ Whenever powered on, participate in a MANET with each node it can hear.

▫ If a computer has access to the Internet then all are able to share

Champaign-Urbana Community

Wireless Network (CUWiN)• A project started in 2000

• Mission

▫ connect more people to Internet and broadband services,

▫ develop open-source hardware and software for use by wireless projects world-wide, and

▫ build and support community-owned, not-for-profit broadband networks in cities and towns around the globe.

• Network architecture

▫ intends to provide a meshed, ad-hoc, non-hierarchical network topology

▫ Routing protocol

� OSPF -> Hazy-Sighted Link State (HSLS) -> Adaptive Hazy-Sighted Link State (A-HSLS)

� scale to thousands or tens of thousands of nodes (they think)

� ETX

Conclusion• Introduction to WMNs

• Network architecture

• Critical design factors

• Advances and research challenges

▫ Scalability

▫ Security

• WMN examples