1. 1 Presented By: Abbas Agane ELG 5125 - University of Ottawa November 29, 2005 Project Presentation Quality of service in ad-hoc networks

2. 2 Agenda Introduction Ad-hoc Network definition Overview: Ad-hoc networks Network architecture Applications of ad-hoc networks Ad-hoc networks characteristics and requirements Overview: QoS What is QoS ? The need of QoS in MANETs Why QoS is hard in MANETs Current Solutions for Support in MANETs Flexible QoS Model for MANETs INSIGNIA-MANETs QoS Signaling Cluster-based Routing Protocol SWAN for MANETs Ad-hoc QoS interconnectivity with Fixed Network Domain services Model for QoS ad-hoc interaction with the host domain Mechanism of operation Ad-hoc QoS interaction with the host domain architecture End-to-end Qos in MANETs connected to Fixed Networks (DS-SWAN) DS-SWAN for upstream Conclusions Q&A

3. 3 Ad Hoc Network definition An ad-hoc network is a wireless LAN, in which some devices are part of the network only for the duration of a communication session or while in some close proximity to the rest of the network. A "mobile ad hoc network" (MANET) is an autonomous system of mobile routers (and associated hosts) connected by wireless links forming an arbitrary graph. Routers are free to move randomly and organize themselves arbitrarily; network topology may change rapidly and unpredictably. May operate in a stand-alone fashion, or may be connected to the Internet. An ad hoc network can be regarded as a “spontaneous network”: a network that automatically “emerges” when nodes gather together

4. 4 MANET – Mobile Ad hoc NETworks A CB D - Mobility- Self configuring and healing- Rapid Deployment - High capacity- Independent of public infrastructure- Relaying - Internet compatible standards-based wireless systems

5. 5 Network Architecture Flat network infrastructure Multi-layered network infrastructure Cluster Head Cluster Head Cluster Head

6. 6 Applications of Ad Hoc Networks  Personal communications cell phones, laptops  Cooperative environments taxi cab network meeting rooms  Emergency operations policing and fire fighting  Military environments Battlefield  Network of sensors or floats over water

7. 7 Ad Hoc Networks Characteristics and Requirements Autonomous and spontaneous nature of nodes Distributed Algorithms to support security, reliability and consistency of exchanged and stored information Time-varying network topology (no pre-existing infrastructure or central administration) Scalable routing and mobility management techniques to face network dynamics Fluctuating link capacity and network resources Enhanced functionalities to improve link layer performance, QoS network support and end-to-end efficiency Low-power devices Energy conserving techniques at all layers

8. 8 What is QoS ? Hard to agree on a common definition of QoS A QoS enabled network shall ensure: That its applications and/or their users have their QoS parameters fulfilled, while at the same time ensuring an efficient resource usage That the most important traffic still has its QoS parameters fulfilled during network overload What are the most important QoS parameters: Throughput, availability, delay, jitter and packet loss

9. 9 The need for QoS in MANETs Applications have special service requirements VoIP: delay, jitter, minimum bandwidth Needs intelligent buffer handling and queueing High mobility of users and network nodes Routing traffic is important No retransmission of lost broadcast messages Routing contol messages must be prioritized For use in emergency and military operations User traffic prioritization is needed user, role, situation etc Wireless bandwidth and battery capacity are scarce resources Need efficient resource usage E.g. only route high priority traffic through terminals that are low on power Need QoS aware routing

10. 10 Why QoS is Hard in Mobile Ad Hoc Networks? Dynamic network topology Flow stop receiving QoS provisions due to path disconnections New paths Must be established, causing data loss and delays Imprecise state information Link state changes continuously Flow states change over time No central control for coordination Error-prone shared medium Hidden terminal problem Limited resources availability Bandwidth, battery life, storage, processing capabilities Insecure medium

11. 11 Current Solutions for QoS support in Mobile Ad Hoc Networks Because of the unique characteristics of the ad-hoc environment three models provide some good insight into the issues of QoS in MANETs These models provide a comprehensive solutions, namely INSIGNIA FQMM SWAN  FQMM  INSIGNIA  SWAN Can be integrated with multiple routing protocols Flexibility!  

12. 12 Flexible QoS Model for MANETs (FQMM) First QoS Model proposed in 2000 for MANETs by Xiao et al Proposes a “hybrid” provisioning that combines the per-flow granularity on IntServ and per-class granularity of DiffServ Adopts DiffServ, but improves the per-class granularity to per-flow granularity for certain class of traffic Built over IntServ and DiffServ models, it can operate with extranet traffic Classification is made at the source node QoS provisioning is made on every node along the path FQMM Model provisions the traffic into two portions the highest priority is assigned per-flow granularity. the rest is assigned per-class granularity. Three types of nodes defined Ingress (transmit) Interior (forward) Egress (receive)

13. 13 INSIGNIA – MANETs QoS Signaling First signaling protocol designed solely for MANETs by Ahn et al. 1998 In-band signaling Base and enhanced QoS levels Per-flow management Resources management adapted as technology Intelligent packet scheduling Flow reservation, restoration and adaptation QoS reports periodically sent to source node Source node takes action to adapt flows to observed network condition Routing Any routing protocol can be used Route maintenance procedure will affect In-band signaling Establish, adapt, tear down reservations Control information embedded in data packets

14. 14 INSIGNIA – OPTION Field Supports in-band signaling by adding a new option field in the IP header to carry the signaling control Reservation Mode (REQ/RES): indicates whether there is already a reservation for this packet. If “no”, the packet is forwarded to INSIGNIA Module which in coordination with a AC may either: grant resources  Service Type = RT (real-time). deny resources  Service Type = BE (best-effort). If “yes”, the packet will be forwarded with the allowed resources. Bandwidth Request (MAX/MIN): indicates the requested amount of bandwidth.

15. 15 INSIGNIA – Bottleneck Node During the flow reservation process a node may be a bottleneck: The service will degrade from RT/MAX -> RT/MIN. If M2 is heavy-loaded it may also degrade the service level to BE/MIN where there is actually no QoS.

16. 16 Cluster-based Routing Protocol for Mobile Ad hoc Networks When network size increase, flat routing schemes become infeasible.  hierarchical routing Explicit hierarchy Group nodes geographically close to each other into explicit clusters Clusterhead Communicate to other nodes on behalf of the cluster Clustering is: a distributed, efficient, scalable protocol Use clustering approach to minimize on-demand route discovery traffic use “local repair” to reduce route acquisition delay and new route discovery traffic suggest a solution to use uni-directional links

17. 17 Cluster Formation Source Destination routing: showing a data path from source to destination

18. 18 Cluster Formation Objective: Form small, stable clusters with only local information  Mechanism:  Variations of “min-id” cluster formation algorithm.  Nodes periodically exchange HELLO pkts to maintain a neighbor table  neighbor status (C_HEAD, C_MEMBER, C_UNDECIDED)  link status (uni-directional link, bi-directional link)  maintain a 2-hop-topology link state table HELLO message format:

19. 19 SWAN Stateless Wireless Ad Hoc Networks An alternative to INSIGNIA with improved scalabilities properties Is a stateless network scheme designed specifically for MANETs with no need to process complex signaling, or to keep per-flow information, to achieve scalability and robustness Promotes rate control system that can be used at each node to treat traffic either as real-time or best-effort Excessive real-time traffic is automatically demoted to best-effort While provides a model that deals with traffic on a per-class, it uses merely two level of service, best-effort and real-time traffic Both level of service can be mapped to DCSPs with known PHB (based on bandwidth requirement) to facilitate extranet QoS May decide to demote part of the real-time traffic to best-effort service due to lack of resources The transmission rate for the best-effort traffic is locally estimated and adjusted to accommodate the bandwidth required by Real Time traffic Supports source-based admission control and distributed congestion control for real-time traffic Uses explicit congestion notification (ECN)

20. 20 ad-hoc QoS interconnectivity with fixed network Ad-Hoc network needs to cling to a host network in order to gain access to the internet Co-operation between ad hoc network and the host network can facilitate end-to-end QoS support Framework proposed by Morgan and Kunz defines a solution for interaction between ad hoc and host networks This framework is not affected by the specific QoS model implemented on either side Ad-Hoc network may decide to implement INSIGNIA, SWAN, or FQMM, while host network may decide to implement DiffServ or IntServ Ad-hoc networks rely on the host network resources and services in order to access to the outside world The host network provides support for the ad-hoc by providing access to specific domain services and agreements Domain services are expressed in terms of three major components

21. 21 Domain services Service Level Agreement (SLA): Fixed networks define SLA as a contract between a customer and service provider that specifies, what services the network service provider will furnish Ad hoc domain: may decide to use any protocol such as SLP (service Location Protocol ) to locate specific services such as a mail server, based on individual needs Traffic Conditioning Agreement (TCA): Specifying classifier rules and any corresponding traffic profiles and metering and shaping rules which are to apply the traffic streams selected by the classifier An example of TCA is the DSCP mapping, and packet fragmentation Ad Hoc network: need to adopt a set of DSCP codes in order to be able to deal with DiffServ QoS traffic Service Provisioning Policy: how traffic conditioners are configured on domain boundary nodes and how traffic streams are mapped to behaviour aggregates to achieve a range of services

22. 22 Model for QoS ad-hoc interaction with host domain Network Elements [1],[2]

23. 23 Mechanism of Operation The GW to the proposed friendly domain can use SLA and TCA proposed by its fixed domain only GW(A’) adopts SLA and TCA proposed by domain DS’ While GW(A”) adopts SLA and TCA proposed by domain DS” The GW has to achieve a compromise between the costs using different services When a GW looses link connectivity during a per-class, extranet packets have to be rerouted to an alternate GW Otherwise it will return to the originating node with a proper error code GWs have to create a table of the in-service DSCP This table provides a way of finding an alternate GW When a GW looses link connectivity during a per-flow session, extranet packets have to be returned to the sender with an error report

24. 24 Aggregate RSVP Is used to solve the scalability issues of RSVP protocol It is particular efficient for inter-domain reservations The terminal ad hoc network is good to employ aRSVP Since, all ad-hoc extranet traffic have to pass through an access network aRSVP is used to configure an aggregate PHB between nodes A’, A”, on one hand and D’, D” on the other hand All end-to-end reservations that use RSVP will use the same aggregate if they belong to the same class All same class reservations will share resources reserved by a single aRSVP This raises the problem of dealing with bursty traffic, because it will simply eat up the resources of other flow Because, Bursty traffic will simply eat up resources of other flows Proved that the performance degradation due to bursty flow comes with performance enhancement in the form of reduction of delay in the tail of the delay distribution

25. 25 Pro-active and reactive approach Proactive approach, by allowing the first or best AN to place an aRSVP request to reserve all classes of traffic (i.e. DSCP) Then other users will use pre-configured services, and only solicit a request for upgrade when needed Problem is the reservation of unused resources in anticipation of future need Unused resources can be released until needed. When needed, they can simply activated Reactive approach, by reserving services only when needed When services for a new DSCP are needed, the GW will broadcast a solicit message requiring all ANs to reply with the level of service and cost they can obtain from a specified host domain GW then will apply a selection criteria to choose which AN should provide aRSVP connection Reactive approach does not reserve unused resources like the proactive one However, a certain delay is expected to find the right AN, and to perform versus reactive aRSVP reservation can be determined from the service policy-provisioning repository

26. 26 Ad-hoc QoS interaction with host domain Architecture Ad-hoc may employ FQMM, SWAN, or INSIGNIA, and may be using dRSVP Ad-hoc will have a traffic forwarding algorithm, which will use the service policies in order to perform QoS routing SLA, TCA, and service provisioning policies, are all imported GW has a common access to SLA, TCA, and service provisioning policies Architecture Elements [1], [2]

27. 27 End-To-End QoS in MANETs Connected to Fixed Networks DS-SWAN (Diff-SWAN) New protocol proposed by Remondo, designed to support end-to- end QoS in ad-hoc networks connected to fixed DiffServ domain DS-SWAN warns nodes in the ad-hoc networks when congestion is excessive for the correct functioning of real-time applications These nodes react by slowing down best-effort traffic DS-SWAN significantly improves end-to-end delays for real-time flows without starvation of background traffic DS-SWAN, the ingress edge router periodically monitors the number of Expedited Forwarding (EF) packets that are dropped by its token bucket meter On the other hand, the corresponding nodes in the fixed IP network periodically monitor the average end-to-end delays of the real-time flows DS-SWAN has been designed to combat the effect of congestion due to excess of best-effort traffic on end-to-end delay real-time flows

28. 28 DS-SWAN for upstream traffic  For Real-time traffic, the DiffServ service class is the Expedite Forwarding PHB (Peer-Hop Behaviour)  The number of dropped packets at the ingress edge router and the end-to-end delay of the real-time connection are associated with the QoS parameters of the SWAN model in the ad hoc network  If the rate of the best-effort leaky bucket traffic shaper is lower, then best- effort traffic is more efficiently restricted and real-time traffic is not so much influenced by best-effort traffic, thereby maintaining the required QoS

29. 29 When a destination node detects that the end-to-end delay of one VoIP flow approached the threshold (i.e. becomes greater than 140ms), it sends a QoS_LOST warning to the ingress edge route When the edge router sends a QoS_LOST to the ad hoc network, it sends the message only to the VoIP sources generating flows that have problems to keep their end-to-end delay under 150ms, which will obviously also arrive at the intermediate nodes along the routes All these nodes forward the QoS_LOST message to all their neighbours because they may be contending with them for medium access DS-SWAN for upstream traffic (cont’)

30. 30 DS-SWAN for upstream traffic (cont’) The nodes in the ad hoc network use priority scheduling at the MAC layer to prioritize routing packets and QoS_LOST packets When a node in ad hoc network receives the QoS_LOST message, it will react by modifying the parameter value in the AIMD rate control algorithm Every time that a QoS_LOST message is received, the node decreases the value of c by ∆c-bit/s with a certain minimum value When no QoS_LOST message is received during T seconds the node increases the value of c by ∆c+bit/s unless the initial value of c has reached For r is opposite of the above results r-> ∆r-bit/s/ ∆c+bit/s

31. 31 Conclusion In this project, I have presented different existing QoS model for wireless ad-hoc networks and a proposed frameworks for ad-hoc interconnectivity with fixed domains INSIGNIA, SWAN, FQMM and DS-SWAN, each model provide the basics for a more comprehensive model Mobile nodes can connect to the Internet gateways of different types, providing different QoS Classified different approach with respect to different mobility scenarios Furthermore, I presented existing classified different level of QoS for hybrid fixed networks In order to achieve an end-to-end QoS approach, QoS information in both fixed and ad-hoc networks should be involved This demands an interaction between the sections

32. 32 References [1] Towards End-to-End QoS in Ad-Hoc Networks Connected to Fixed Networks David Remondo Catalonia Univ. of Technology (UPC) [2] An architectural framework for MANET QoS interaction with access domains Yasser Morgan and Thomas Kunz, Carleton University [3]A proposal for an ad-hoc network QoS gateway Yasser Morgan and Thomas Kunz, Carleton University [4] A Glance at Quality of Services in Mobile Ad-Hoc Networks Zeinalipour-Yazti Demetrios (csyiazti@cs.ucr.edu)csyiazti@cs.ucr.edu [5] Quality of Service in Ad-Hoc Networks Eric Chi, Antoins Dimakis el (smartnets@uclink.berkeley.edu) [6] QoS in Mobile Ad Hoc Networks Prasant Mohapatra, Jian Li and Chao Gui, University of California [7] QoS-aware Routing Based on Bandwidth Estimation for Mobile Ad Hoc networks Lei Chen and Wendi Heinzelman, University of Rochester{chenlei, wheinzel}@ece.rochester.edu [8] Dynamic Quality of Service for Mobile Ad-Hoc Networks M. Mirhakkak, N. Schult, D. Thomson, The MITRE Corporation [9] Network Architecture to Support QoS in Mobile Ad Hoc Networks Lei Chen and Wendi Heizelman, University of Rochester

33. 33 Thank You! Questions? Q&A