1. WIRELESS MESH NETWORKS Ian F. AKYILDIZ* and Xudong WANG** *Georgia Institute of Technology BWN (Broadband Wireless Networking) Lab & **TeraNovi Technologies

2. 2 7. NETWORK CONTROL AND MANAGEMENT MANAGEMENT 2

3. 3 Introduction In a traditional wired network, the goal of a network management system is to assist network managers in monitoring and maintaining network via some tools, applications, or devices. In a traditional wired network, the goal of a network management system is to assist network managers in monitoring and maintaining network via some tools, applications, or devices. e.g., Simple Network Management Protocol (SNMP) and Common Management Information Protocol (CMIP). 3

4. 4 4 Introduction In these protocols, network management consists of the following major functions: * Performance Management * Configuration Management * Accounting Management * Fault Management * Security Management

5. 5 5 Performance Management Maintains the network performance at an acceptable level by monitoring network behavior, measuring various parameters of the network, and taking certain actions when necessary.

6. 6 6 Configuration Management Network and system configuration information is monitored so that operation on different version of network hardware and software can be managed or tracked.

7. 7 7 Accounting Management Network utilization is measured in order to appropriately regulate network use by users.

8. 8 8 Fault Management This function detects network problems and then log them or notify them to users. Moreover, some capabilities of automatically fixing problems are expected.

9. 9 9 Security Management This function provides access control in order to avoid network access by unauthorized users.

10. 10 Mobility Management For WMNs, more functions need to be available in the network management For WMNs, more functions need to be available in the network management For example, mobility management For example, mobility management 10

11. 11 Topology Control and Management Topology in WMNs can be variable because of Topology in WMNs can be variable because of * node mobility * node mobility * node association/disassociation, and * node association/disassociation, and * fluctuating signal quality in links, * fluctuating signal quality in links, it needs to be controlled so as to stabilize the network and it needs to be controlled so as to stabilize the network and achieve a better performance. achieve a better performance. Thus, topology control and management are critical for WMNs. Thus, topology control and management are critical for WMNs. 11

12. 12 POWER MANAGEMENT The power level of a node in WMNs closely impacts The power level of a node in WMNs closely impacts * the network topology, * the network topology, * the transmission rate of a node, and * the transmission rate of a node, and * also the energy efficiency, * also the energy efficiency, so the power management is another critical function for so the power management is another critical function for network management. network management. 12

13. 13 TIMING SYNCHRONIZATION To assist communication protocols, sometimes timing To assist communication protocols, sometimes timing synchronization is needed among nodes. synchronization is needed among nodes. e.g., any protocol based on the concept of e.g., any protocol based on the concept of TDMA must rely on synchronized timing among TDMA must rely on synchronized timing among different nodes. different nodes. These additional functions highlight new issues of These additional functions highlight new issues of network management in WMNs. network management in WMNs. 13

14. 14 Mobility Management Mobility management consists of two important tasks: Mobility management consists of two important tasks: location and handoff management location and handoff management Location management  location registration and call delivery, Location management  location registration and call delivery, Handoff management  handoff initiation, new connection generation, and data flow control for call handoff. Handoff management  handoff initiation, new connection generation, and data flow control for call handoff. 14

15. 15 Mobility Management Several features of WMNs make their mobility management much different from that of both cellular networks and mobile ad hoc networks. Several features of WMNs make their mobility management much different from that of both cellular networks and mobile ad hoc networks. In a cellular network, from client to the network there is always one wireless hop. In a cellular network, from client to the network there is always one wireless hop. However, in WMNs there exist multiple wireless hops from mesh client to the Internet. However, in WMNs there exist multiple wireless hops from mesh client to the Internet. 15

16. 16 Mobility Management In some mobile ad hoc networks, Internet access is not a concern, so user mobility can be handled by link layer handoff and routing protocol. In some mobile ad hoc networks, Internet access is not a concern, so user mobility can be handled by link layer handoff and routing protocol. In other words, mobility management can be very simple or even not necessary In other words, mobility management can be very simple or even not necessary In a mobile ad hoc network that requires Internet access, the complexity of mobility management increases. In a mobile ad hoc network that requires Internet access, the complexity of mobility management increases. 16

17. 17 Mobility Management However, network nodes do not need to provide network access for other network nodes, which can simplify the mobility management architecture. However, network nodes do not need to provide network access for other network nodes, which can simplify the mobility management architecture. The above differences highlight the challenges in mobility management for WMNs. The above differences highlight the challenges in mobility management for WMNs. 17

18. 18 Mobility Management However, compared to mobile ad hoc networks, WMNs However, compared to mobile ad hoc networks, WMNs have their own advantages such as mesh routers are have their own advantages such as mesh routers are usually stationary and do not have a constraint on usually stationary and do not have a constraint on energy and processing power. energy and processing power. Taking advantage of these features is helpful to reduce the complexity of mobility management for WMNs. Taking advantage of these features is helpful to reduce the complexity of mobility management for WMNs. 18

19. 19 Mobility Management Although standards are available for WMNs, mobility management is still a least-specified area. Although standards are available for WMNs, mobility management is still a least-specified area. Mobility management of IEEE 802.11 based WMNs is not fully investigated by the 802.11s task group (IEEE 802.11s 2006) and the interim solution is based on IEEE 802.11f roaming solution (IEEE802.11f 2003). Mobility management of IEEE 802.11 based WMNs is not fully investigated by the 802.11s task group (IEEE 802.11s 2006) and the interim solution is based on IEEE 802.11f roaming solution (IEEE802.11f 2003). 19

20. 20 Mobility Management However, this protocol was designed for the roaming from one access point to another, and has the drawback of a very large latency. However, this protocol was designed for the roaming from one access point to another, and has the drawback of a very large latency. Therefore, IEEE 802.11f itself is not active now, and roaming of IEEE networks is currently specified in IEEE 802.11r (IEEE 802.11r 2006). Therefore, IEEE 802.11f itself is not active now, and roaming of IEEE networks is currently specified in IEEE 802.11r (IEEE 802.11r 2006). 20

21. 21 Mobility Management Neither IEEE 802.11f nor IEEE 802.11r is applicable to IEEE 802.11s due to significant differences in the network architecture. Neither IEEE 802.11f nor IEEE 802.11r is applicable to IEEE 802.11s due to significant differences in the network architecture. So far, mobility management of IEEE 802.16 WMNs has not been specified in all available standards and drafts (IEEE 802.16 2004), (IEEE 802.16a 2003), (IEEE 802.16e 2005). So far, mobility management of IEEE 802.16 WMNs has not been specified in all available standards and drafts (IEEE 802.16 2004), (IEEE 802.16a 2003), (IEEE 802.16e 2005). 21

22. 22 Mobility Management Initially, IEEE 802.16 was designed for fixed wireless access. Initially, IEEE 802.16 was designed for fixed wireless access. In order to support mobile clients, an extended version, IEEE 802.16e, was proposed (IEEE 802.16e 2005). In order to support mobile clients, an extended version, IEEE 802.16e, was proposed (IEEE 802.16e 2005). However, the IEEE 802.16e standard only cover mobility support for the point-to-multipoint (PMP) mode. However, the IEEE 802.16e standard only cover mobility support for the point-to-multipoint (PMP) mode. 22

23. 23 The specific features of WMNs make research results of mobility management obtained for other wireless networks be not applicable to WMNs. The specific features of WMNs make research results of mobility management obtained for other wireless networks be not applicable to WMNs. On the other hand, some concepts from cellular, On the other hand, some concepts from cellular, mobile IP, or mobile ad hoc networks can be borrowed mobile IP, or mobile ad hoc networks can be borrowed for WMNs. for WMNs. 23 Mobility Management

24. 24 So far among the limited number of research papers on mobility management of WMNs, none of them has really investigated the following issues [1]: So far among the limited number of research papers on mobility management of WMNs, none of them has really investigated the following issues [1]: For the system architecture, what are the key For the system architecture, what are the key differences between mobility management of WMNs differences between mobility management of WMNs and that of cellular networks and mobile IP? and that of cellular networks and mobile IP? 24 Mobility Management [1] J. Xie and X. Wang, “A Survey of Mobility Management in Hybrid Wireless Mesh Networks,’’ IEEE Network Magazine, vol. 22, no. 6, pp. 34-40, Nov.-Dec. 2008.

25. 25 Mobility Management What is the best system architecture for the mobility management of WMNs? What is the best system architecture for the mobility management of WMNs? What are the challenging issues that are specific to the mobility management of WMNs? What are the challenging issues that are specific to the mobility management of WMNs? 25

26. 26 Mobility Management Existing mobility management schemes for WMNs usually follow the same system architecture as that of mobile IP or cellular networks Also inherit similar procedures of location management and handoff mechanisms from cellular networks or mobile IP. 26

27. 27 Mobility Management * However, the overall system architecture and operation procedures of a mobility management protocol for WMNs must be developed by following a novel framework due to the specific features pertained by WMNs. 27

28. 28 Architecture of Mobile Mesh IP In general, mobility exists in either mesh clients or mesh routers. In general, mobility exists in either mesh clients or mesh routers. For example, considering a localized mesh network on the train, both mesh routers and mesh clients are mobile relative to the WMN on the ground. For example, considering a localized mesh network on the train, both mesh routers and mesh clients are mobile relative to the WMN on the ground. 28

29. 29 Architecture of Mobile Mesh IP However, it is also true that in most applications mesh routers in WMNs remain stationary. However, it is also true that in most applications mesh routers in WMNs remain stationary. Thus, in mobile mesh IP, we focus on the mobility management for mesh clients. Thus, in mobile mesh IP, we focus on the mobility management for mesh clients. Moreover, we assume no ad hoc networking between mesh clients is necessary. Moreover, we assume no ad hoc networking between mesh clients is necessary. 29

30. 30 Architecture of Mobile Mesh IP The concept of mobile mesh IP is depicted in Fig. 7.1. The concept of mobile mesh IP is depicted in Fig. 7.1. The mobile mesh IP is compatible with mobile IP since the concept of HA and FA in mobile IP are adopted. The mobile mesh IP is compatible with mobile IP since the concept of HA and FA in mobile IP are adopted. However, due to the WMN environment, the protocol architecture, mechanism, and operation procedures of mobile mesh IP are much different from those of mobile IP. However, due to the WMN environment, the protocol architecture, mechanism, and operation procedures of mobile mesh IP are much different from those of mobile IP. 30

31. 31 Procedure of Mobile Mesh IP and Remaining Work 31

32. 32 Architecture of Mobile Mesh IP The entire network consists of three hierarchical layers: Internet backbone, mesh backbone, and mesh clients. The entire network consists of three hierarchical layers: Internet backbone, mesh backbone, and mesh clients. Mesh backbone is comprised of localized mesh networks distributed in different geographical areas. Mesh backbone is comprised of localized mesh networks distributed in different geographical areas. 32

33. 33 Architecture of Mobile Mesh IP We define such a localized mesh network as a mesh domain. We define such a localized mesh network as a mesh domain. Mesh backbone is connected to the Internet via wired links between mesh routers and access routers (AR). Mesh backbone is connected to the Internet via wired links between mesh routers and access routers (AR). A mesh router connected to an AR is usually called a mesh gateway. A mesh router connected to an AR is usually called a mesh gateway. 33

34. 34 Architecture of Mobile Mesh IP In general, in one mesh domain only a small number of mesh routers are connected to their ARs. In general, in one mesh domain only a small number of mesh routers are connected to their ARs. In the simplest case, only one mesh gateway is connected to its corresponding AR. In the simplest case, only one mesh gateway is connected to its corresponding AR. However, if capacity is the concern, then multiple mesh gateways may be needed and are connected to their own ARs. However, if capacity is the concern, then multiple mesh gateways may be needed and are connected to their own ARs. 34

35. 35 Architecture of Mobile Mesh IP Similar to mobile IP, mobile mesh IP also needs home agent (HA) and foreign agent (FA) to support mobility management. Similar to mobile IP, mobile mesh IP also needs home agent (HA) and foreign agent (FA) to support mobility management. HA is located at the Internet backbone, and can be distant from WMNs. HA is located at the Internet backbone, and can be distant from WMNs. 35

36. 36 Architecture of Mobile Mesh IP An FA can be located together with an AR, a mesh gateway, or even a regular mesh router, depending on the design. An FA can be located together with an AR, a mesh gateway, or even a regular mesh router, depending on the design. Accordingly, the care-of-address (CoA) of a mesh client can be handled at an AR, a mesh gateway, or a regular mesh router. Accordingly, the care-of-address (CoA) of a mesh client can be handled at an AR, a mesh gateway, or a regular mesh router. 36

37. 37 Procedure of Mobile Mesh IP and Remaining Work Mobile mesh IP is explained below. Mobile mesh IP is explained below. For simplicity, we assume the FA is located with the For simplicity, we assume the FA is located with the mesh router that has direct connection to AR. mesh router that has direct connection to AR. However, the optimal location of the FA is subject However, the optimal location of the FA is subject to research. to research. 37

38. 38 Procedure of Mobile Mesh IP: Mesh client moves within the same mesh domain If a mesh client does not change its attached mesh router, If a mesh client does not change its attached mesh router, then nothing needs to be done. then nothing needs to be done. When the mesh client detects that its attached mesh router is about to change, layer-2 handoff is invoked. When the mesh client detects that its attached mesh router is about to change, layer-2 handoff is invoked. After handoff is done, packets will be routed to the mesh client via the new mesh router. After handoff is done, packets will be routed to the mesh client via the new mesh router. 38

39. 39 Procedure of Mobile Mesh IP This re-routing process can be handled directly by the mesh routing protocol. This re-routing process can be handled directly by the mesh routing protocol. However, in order to avoid packet loss and reduce re- routing delay, route re-discovery may need to be done before handoff is complete. However, in order to avoid packet loss and reduce re- routing delay, route re-discovery may need to be done before handoff is complete. It should be noted that route re-discovery only occurs in the same mesh domain, and in all cases no location update is needed. It should be noted that route re-discovery only occurs in the same mesh domain, and in all cases no location update is needed. 39

40. 40 Procedure of Mobile Mesh IP: Mesh client moves from one mesh domain to another The change of mesh domain results in the change of CoA. The change of mesh domain results in the change of CoA. Thus, this scenario is much more complicated than the previous scenario. Thus, this scenario is much more complicated than the previous scenario. First of all, layer-2 handoff is needed. First of all, layer-2 handoff is needed. Secondly, location update needs to be performed to locate the new FA for the mesh client, update the CoA at the HA, and find a new routing path from the HA to the FA. Secondly, location update needs to be performed to locate the new FA for the mesh client, update the CoA at the HA, and find a new routing path from the HA to the FA. Thirdly, when location update is done, the mesh routing protocol needs to find a new routing path from the new FA to the mesh client Thirdly, when location update is done, the mesh routing protocol needs to find a new routing path from the new FA to the mesh client 40

41. 41 Mobile Mesh IP: Remaining Work Many problems in mobile mesh IP are different from mobile IP and thus require research efforts to solve. Many problems in mobile mesh IP are different from mobile IP and thus require research efforts to solve. Overall Architecture Design: The major task in architecture design is to determine where an FA should be located. Overall Architecture Design: The major task in architecture design is to determine where an FA should be located. 41

42. 42 If it is with an AR or a mesh gateway, then a mesh If it is with an AR or a mesh gateway, then a mesh client is difficult to detect the change of CoA, and client is difficult to detect the change of CoA, and thus a fast and efficient scheme is needed to detect thus a fast and efficient scheme is needed to detect the change of CoA. the change of CoA. On the other hand, if an FA is with a regular mesh On the other hand, if an FA is with a regular mesh router, then it is easy for a mesh client to detect the router, then it is easy for a mesh client to detect the change of CoA. change of CoA. 42 Mobile Mesh IP: Remaining Work

43. 43 However, there exist several issues. However, there exist several issues. Firstly, it is difficult to maintain an FA because of Firstly, it is difficult to maintain an FA because of being separated from AR by multiple wireless hops. being separated from AR by multiple wireless hops. Secondly, a mesh client may experience too frequent change Secondly, a mesh client may experience too frequent change of CoA, unless change of the attached mesh router does not of CoA, unless change of the attached mesh router does not mean a change of CoA. mean a change of CoA. 43 Mobile Mesh IP: Remaining Work

44. 44 In the latter case, coordination between mesh routers are needed. In the latter case, coordination between mesh routers are needed. In either case, signaling messages of updating of CoA have to traverse a multihop wireless network, which can experience large delay and cause a large percentage of overhead. In either case, signaling messages of updating of CoA have to traverse a multihop wireless network, which can experience large delay and cause a large percentage of overhead. 44 Mobile Mesh IP: Remaining Work

45. 45 Fast route re-discovery: Fast route re-discovery: This is required no matter whether location This is required no matter whether location update happens. update happens. When location update is not needed, route re- When location update is not needed, route re- discovery only occurs in the mesh routing discovery only occurs in the mesh routing protocol. protocol. 45 Mobile Mesh IP: Remaining Work

46. 46 Thus, mesh routing protocol needs to be quick in Thus, mesh routing protocol needs to be quick in route re-discovery. route re-discovery. When location update is needed, additional When location update is needed, additional efforts are needed to quickly find a new routing efforts are needed to quickly find a new routing path from HA to the new FA in the WMN. path from HA to the new FA in the WMN. 46 Mobile Mesh IP: Remaining Work

47. 47 Fast Location Update: Fast Location Update: In addition to fast route re-discovery, the location In addition to fast route re-discovery, the location update procedure must be designed to quickly locate update procedure must be designed to quickly locate the new FA for the mesh client. the new FA for the mesh client. Since multi-hop networks exist between the FA and Since multi-hop networks exist between the FA and the mesh client, locating a new FA is not obvious, and the mesh client, locating a new FA is not obvious, and thus demands an efficient scheme. thus demands an efficient scheme. 47 Mobile Mesh IP: Remaining Work

48. 48 This task is correlated with the detection of CoA This task is correlated with the detection of CoA change. change. When CoA change is detected, the first step of location update is to find the new FA. When CoA change is detected, the first step of location update is to find the new FA. Once FA is found, the CoA of the mesh client is updated at the HA. Once FA is found, the CoA of the mesh client is updated at the HA. 48 Mobile Mesh IP: Remaining Work

49. 49 Direct mesh routing across different mesh domains: When a mesh router can be connected to other mesh routers When a mesh router can be connected to other mesh routers in different mesh domains, it may forward packets from one in different mesh domains, it may forward packets from one mesh domain directly to another one for the purpose of load mesh domain directly to another one for the purpose of load balancing. balancing. In this case, the mobility management becomes really In this case, the mobility management becomes really challenging, because this mesh router needs to be associated challenging, because this mesh router needs to be associated with several gateways with different IP subnets. with several gateways with different IP subnets. 49 Mobile Mesh IP: Remaining Work

50. 50 When a mesh client is associated with this mesh router during roaming, a new scheme is needed for the mesh router to decide what subnet shall be used to allocate an IP address for the mesh client. When a mesh client is associated with this mesh router during roaming, a new scheme is needed for the mesh router to decide what subnet shall be used to allocate an IP address for the mesh client. This issue reflects the conflict between IP routing and mesh routing. This issue reflects the conflict between IP routing and mesh routing. 50 Mobile Mesh IP: Remaining Work

51. 51 Cross-layer design: Due the multi-hop wireless networks in WMNs, cross- layer design between a mobility management protocol and MAC and routing protocols is critical. Due the multi-hop wireless networks in WMNs, cross- layer design between a mobility management protocol and MAC and routing protocols is critical. For example, signaling messages of location management have to traverse a multihop wireless network. For example, signaling messages of location management have to traverse a multihop wireless network. 51 Mobile Mesh IP: Remaining Work

52. 52 To improve the performance of location management, MAC and routing need to be improved to reduce the delay of signaling messages and increase the delivery ratio of such messages. To improve the performance of location management, MAC and routing need to be improved to reduce the delay of signaling messages and increase the delivery ratio of such messages. 52 Mobile Mesh IP: Remaining Work

53. 53 Open Research Issues * New framework of mobility management: Mobile mesh IP is an example to illustrate the new framework of mobility management ofWMNs. Mobile mesh IP is an example to illustrate the new framework of mobility management ofWMNs. However, it still contains many issues to be resolved. However, it still contains many issues to be resolved. Moreover, schemes with better architecture and design than mobile mesh IP also need to be developed. Moreover, schemes with better architecture and design than mobile mesh IP also need to be developed. 53

54. 54 Open Research Issues * Support of hierarchical mesh backbone: Mesh backbone may consist of heterogeneous mesh routers organized hierarchically. Mesh backbone may consist of heterogeneous mesh routers organized hierarchically. Under this architecture, mobility management becomes more complicated, since a hierarchical wireless network is involved in all procedures such as handoff, location update, and route re-discovery. Under this architecture, mobility management becomes more complicated, since a hierarchical wireless network is involved in all procedures such as handoff, location update, and route re-discovery. 54

55. 55 Open Research Issues * Multi-layer schemes: Mobility management is closely related to multiple layers of network protocols, the development of multi-layer mobility management schemes is an interesting topic. Mobility management is closely related to multiple layers of network protocols, the development of multi-layer mobility management schemes is an interesting topic. 55

56. 56 Open Research Issues * Cross-layer design: To improve the overall performance of mobility management, cross-layer design plays an important role. To improve the overall performance of mobility management, cross-layer design plays an important role. For example, many functions such as fast handoff and fast router-rediscovery all depend on cooperative work between multiple layers. For example, many functions such as fast handoff and fast router-rediscovery all depend on cooperative work between multiple layers. 56

57. 57 New Location Services A desired feature by WMNs. A desired feature by WMNs. Location information can enhance the performance of Location information can enhance the performance of MAC and routing protocols. MAC and routing protocols. It can help to develop promising location-related It can help to develop promising location-related applications. applications. Proposing accurate or efficient algorithms for Proposing accurate or efficient algorithms for location service is still an open research topic. location service is still an open research topic. 57

58. 58 Power Management Power level is an important parameter that can Power level is an important parameter that can impact a wireless network in different ways. impact a wireless network in different ways. Thus, control the power level or dynamic adjustment Thus, control the power level or dynamic adjustment of power level is critical to all wireless networks. of power level is critical to all wireless networks. 58

59. 59 Power Management Usually power control aims to determine the Usually power control aims to determine the transmit power level such that the received power transmit power level such that the received power is maintained at a desire level no matter what is maintained at a desire level no matter what channel characteristics and what distance between channel characteristics and what distance between transmitter and receiver. transmitter and receiver. However, power management is concerned with However, power management is concerned with more generic target. more generic target. For example, power level needs to be controlled to save For example, power level needs to be controlled to save energy consumption, improve topology, and so on. energy consumption, improve topology, and so on. 59

60. 60 Key Aspects of Power Management * Save energy consumption * Save energy consumption * Improve network stability * Improve network stability * Adjust transmission rate * Adjust transmission rate * Change network topology * Change network topology 60

61. 61 Power Management in IEEE 802.11 Both power control and management specified in the base Both power control and management specified in the base standard (IEEE 802.11 1999). standard (IEEE 802.11 1999). Power control function provides an option for the MAC Power control function provides an option for the MAC layer to adjust transmit power for each packet as layer to adjust transmit power for each packet as needed. needed. Power management defined in IEEE 802.11 aims to Power management defined in IEEE 802.11 aims to reduce power consumptions in mobile stations (IEEE reduce power consumptions in mobile stations (IEEE 802.11 1999). 802.11 1999). 61

62. 62 Power Management in IEEE 802.11 Mechanisms of power management are specified for both the infrastructure network and the independent basic service set (IBSS). Mechanisms of power management are specified for both the infrastructure network and the independent basic service set (IBSS). In the infrastructure network, power management is controlled by the access point. In the infrastructure network, power management is controlled by the access point. Thus, when a mobile station wants to work in power save Thus, when a mobile station wants to work in power save mode, it needs to inform AP of this request by setting the power management bits in the frame control field of a frame. mode, it needs to inform AP of this request by setting the power management bits in the frame control field of a frame. 62

63. 63 Power Management in IEEE 802.11 Once the AP is notified of this operation, it cannot send Once the AP is notified of this operation, it cannot send a packet to the mobile station at any arbitrary time. a packet to the mobile station at any arbitrary time. Instead, it shall buffer packets and send them to the Instead, it shall buffer packets and send them to the mobile station at designated times when the mobile mobile station at designated times when the mobile station is waken from the power save mode. station is waken from the power save mode. 63

64. 64 Power Management in IEEE 802.11 In an IBSS, the principle of power management is similar In an IBSS, the principle of power management is similar to that in the infrastructure network. to that in the infrastructure network. However, power management process is not totally However, power management process is not totally controlled by access points, but is handled by cooperation controlled by access points, but is handled by cooperation between nodes. between nodes. When a mobile station wants to be in PS mode, it has to When a mobile station wants to be in PS mode, it has to inform all other mobile stations of this operation by inform all other mobile stations of this operation by sending an ad hoc traffic indication message (ATIM). sending an ad hoc traffic indication message (ATIM). Thus, when another mobile station has packets to this Thus, when another mobile station has packets to this station in power save mode, it has to buffer packets. station in power save mode, it has to buffer packets. 64

65. 65 Power Management in IEEE 802.11 Different from the infrastructure mode, the mobile Different from the infrastructure mode, the mobile station does not have a designated time to send packets station does not have a designated time to send packets to the mobile station in power save mode. to the mobile station in power save mode. Thus, before it starts to send packets, it has to send an ATIM message Thus, before it starts to send packets, it has to send an ATIM message During ATIM window, all mobile stations are awake, During ATIM window, all mobile stations are awake, so the mobile station can receive such a message, and then wake up to receive packets. so the mobile station can receive such a message, and then wake up to receive packets. 65

66. 66 Power Management in IEEE 802.11 The operation of IEEE 802.11e (IEEE 802.11e 2005), The operation of IEEE 802.11e (IEEE 802.11e 2005), the QoS standard for IEEE 802.11, is different from the QoS standard for IEEE 802.11, is different from the base IEEE 802.11, so some modifications are made the base IEEE 802.11, so some modifications are made to power management in IEEE 802.11e. to power management in IEEE 802.11e. E.g., contention access period (CAP), E.g., contention access period (CAP), contention period (CP), contention period (CP), contention free period (CFP), contention free period (CFP), extended distributed channel access (EDCA), and extended distributed channel access (EDCA), and transmission opportunities (TXOP) transmission opportunities (TXOP) in an IEEE 802.11e frame. in an IEEE 802.11e frame. 66

67. 67 Power Management in IEEE 802.11e Thus, power saving operations need to be changed to adapt to this different frame structure in both the infrastructure network and the IBSS. Thus, power saving operations need to be changed to adapt to this different frame structure in both the infrastructure network and the IBSS. However, the basic power control mechanism of IEEE 802.11e is still the same as that defined in the base standard. However, the basic power control mechanism of IEEE 802.11e is still the same as that defined in the base standard. 67

68. 68 Power Management in IEEE 802.15 In IEEE 802.15, the situation is a little different from IEEE 802.11. In IEEE 802.15, the situation is a little different from IEEE 802.11. In IEEE 802.15.1 (IEEE 802.15.1 2002), power management is defined as a function of the link manager layer. In IEEE 802.15.1 (IEEE 802.15.1 2002), power management is defined as a function of the link manager layer. However, no specific scheme is specified. Power control is specified in IEEE 802.15.1 to provide a power controlled link for devices that have such a capability. However, no specific scheme is specified. Power control is specified in IEEE 802.15.1 to provide a power controlled link for devices that have such a capability. 68

69. 69 Power Management in IEEE 802.15 For IEEE 802.15.4 (IEEE 802.15.4 2003), since it For IEEE 802.15.4 (IEEE 802.15.4 2003), since it targets at low-rate wireless PANs such as sensor nodes, targets at low-rate wireless PANs such as sensor nodes, power control is not considered as a necessary function. power control is not considered as a necessary function. Power management is required according to the standard, Power management is required according to the standard, but the specification of a detailed scheme is out of the but the specification of a detailed scheme is out of the scope of IEEE 802.15.4. scope of IEEE 802.15.4. In IEEE 802.15.3 (IEEE 802.15.3 2003), both power In IEEE 802.15.3 (IEEE 802.15.3 2003), both power control and power management have been specified. control and power management have been specified. 69

70. 70 Power Management in IEEE 802.15 There are two forms of power control for IEEE 802.15.3: There are two forms of power control for IEEE 802.15.3: –Maximum transmit power for contention access period (CAP), beacon, and directed management channel time allocations (MCTAs), –Adjustable power in channel time allocation (CTA). Maximum transmit power control aims to prevent any device in CAP having better access to the medium. Maximum transmit power control aims to prevent any device in CAP having better access to the medium. Adjustable transmit power control is intended to reduce both power usage and interference (IEEE 802.15.3 2003). Adjustable transmit power control is intended to reduce both power usage and interference (IEEE 802.15.3 2003). 70

71. 71 Power Management in IEEE 802.15 The goal of power management in IEEE 802.15.3 is to enable longer operation time for battery powered devices. The goal of power management in IEEE 802.15.3 is to enable longer operation time for battery powered devices. Three techniques are specified to turn off a device for one or more super frames: Three techniques are specified to turn off a device for one or more super frames: –device synchronized power save (DSPS) mode –piconet synchronized power save (PSPS) mode –asynchronous power save (APS) mode. 71

72. 72 Power Management in IEEE 802.15 Thus, a device in a piconet can be in four states: Thus, a device in a piconet can be in four states: –ACTIVE mode –DSPS mode –PSPS mode –APS mode. As usual, IEEE 802.15.3 does not give detailed scheme As usual, IEEE 802.15.3 does not give detailed scheme to utilize the three power save modes to reduce power to utilize the three power save modes to reduce power consumption. consumption. 72

73. 73 Power Management in IEEE 802.16 Power control is specified to determine transmit Power control is specified to determine transmit power level in both uplink and downlink. power level in both uplink and downlink. Power control can be closed or open loop. Power control can be closed or open loop. However, other functionalities of power management However, other functionalities of power management are not specified. are not specified. –For example, how to achieve power efficiency for clients is not defined in IEEE 802.16. 73

74. 74 Power Management in Wireless Mesh Networks Power management in most other wireless networks are focused on power control and power saving operations. Power management in most other wireless networks are focused on power control and power saving operations. However, the objective of having power management in WMNs is much broader. However, the objective of having power management in WMNs is much broader. Moreover, the goal of power management for WMNs varies, and is also different between mesh routers and mesh clients. Moreover, the goal of power management for WMNs varies, and is also different between mesh routers and mesh clients. 74

75. 75 Power Management for Mesh Routers Since mesh routers do not have a constraint on power consumption, power management usually aims to control connectivity, interference, spectrum spatial reuse, topology Since mesh routers do not have a constraint on power consumption, power management usually aims to control connectivity, interference, spectrum spatial reuse, topology The major functionalities that can be achieved through power management are summarized as follows. The major functionalities that can be achieved through power management are summarized as follows. 75

76. 76 Power Management for Mesh Routers: Self-Organize the Network When mesh routers are deployed as a backbone network, their locations are an important factor that impacts the network performance, especially for in-door environment. When mesh routers are deployed as a backbone network, their locations are an important factor that impacts the network performance, especially for in-door environment. Using sophisticated RF site survey may help to find satisfactory locations. Using sophisticated RF site survey may help to find satisfactory locations. - This actually contradicts with the initial motivation of using WMNs, i.e., convenience - This actually contradicts with the initial motivation of using WMNs, i.e., convenience 76

77. 77 Moreover, various deployment factors can impact the Moreover, various deployment factors can impact the performance of WMNs performance of WMNs - Although investigating such factors to find out guidelines for - Although investigating such factors to find out guidelines for deployment of WMNs is helpful to improve network deployment of WMNs is helpful to improve network performance, it is not enough for performance optimization performance, it is not enough for performance optimization considering various contradicting parameters. considering various contradicting parameters. In addition, practically mesh routers should be In addition, practically mesh routers should be deployed by anyone without RF engineering experience. deployed by anyone without RF engineering experience. 77 Power Management for Mesh Routers: Self-Organize the Network Power Management for Mesh Routers: Self-Organize the Network

78. 78 To avoid this dilemma, we can rely on power management to self organize the network in the following way. To avoid this dilemma, we can rely on power management to self organize the network in the following way. First, mesh routers can be deployed by using simple rules First, mesh routers can be deployed by using simple rules For example, in a home environment, users can put mesh routers in any locations as long as routers have approximately same distance in-between. For example, in a home environment, users can put mesh routers in any locations as long as routers have approximately same distance in-between. 78 Power Management for Mesh Routers: Self-Organize the Network Power Management for Mesh Routers: Self-Organize the Network

79. 79 Due to different channel fading and attenuation, the received power level at different routers will be different. Due to different channel fading and attenuation, the received power level at different routers will be different. Thus, the second step is to apply a power management scheme to automatically adjust the power level on each mesh router. Thus, the second step is to apply a power management scheme to automatically adjust the power level on each mesh router. By doing so, mesh routers are organized to form a network that achieve best performance in terms of throughput or other performance metrics By doing so, mesh routers are organized to form a network that achieve best performance in terms of throughput or other performance metrics 79 Power Management for Mesh Routers: Self-Organize the Network Power Management for Mesh Routers: Self-Organize the Network

80. 80 Improve Network Connectivity and Stability This functionality can be viewed as an additional task of network self-organization or a new task if self- organization is not available. This functionality can be viewed as an additional task of network self-organization or a new task if self- organization is not available. It aims to ensure the connectivity or topology of the network is stable It aims to ensure the connectivity or topology of the network is stable 80

81. 81 Improve Network Connectivity and Stability In a wireless mesh network, without power management, link quality between nodes varies from time to time. In a wireless mesh network, without power management, link quality between nodes varies from time to time.  cause some nodes loose connectivity or change their associations with other nodes  cause some nodes loose connectivity or change their associations with other nodes  Such dynamics will finally result in a instable network topology, and severely degrade the performance of MAC, routing, and even transport layer protocols.  Such dynamics will finally result in a instable network topology, and severely degrade the performance of MAC, routing, and even transport layer protocols. Therefore, power levels on mesh routers need to be adjusted dynamically instead of being set up with a fixed number, so that the link quality between nodes is always to maintain at an acceptable level. Therefore, power levels on mesh routers need to be adjusted dynamically instead of being set up with a fixed number, so that the link quality between nodes is always to maintain at an acceptable level. 81

82. 82 Reduce Interference and Increase Spectrum Spatial-reuse Spectrum spatial-reuse is a critical factor that impacts Spectrum spatial-reuse is a critical factor that impacts the scalability of WMNs. the scalability of WMNs. If spectrum spatial-reuse is poor, throughput drops If spectrum spatial-reuse is poor, throughput drops quickly as the number of hops increases, and also the quickly as the number of hops increases, and also the number of users that can be supported is small. number of users that can be supported is small. When a single channel is used in each network node, When a single channel is used in each network node, interference among nodes greatly impacts the spectrum interference among nodes greatly impacts the spectrum spatial-reuse factor. spatial-reuse factor. 82

83. 83 Reduce Interference and Increase Spectrum Spatial-reuse When multiple channels are used, this problem is less severe. When multiple channels are used, this problem is less severe. However, considering one specific channel, its spectrum spatial-reuse is still low. However, considering one specific channel, its spectrum spatial-reuse is still low. 83

84. 84 Reduce Interference and Increase Spectrum Spatial-reuse Thus, for both single-channel or multi-channel mesh Thus, for both single-channel or multi-channel mesh routers, power management is needed to ensure that the routers, power management is needed to ensure that the received power level at each mesh router does not exceed received power level at each mesh router does not exceed a desired level. a desired level. Otherwise, interference will result in a low spatial-reuse Otherwise, interference will result in a low spatial-reuse efficiency efficiency 84

85. 85 Reduce Interference and Increase Spectrum Spatial-reuse However, lower received power levels will cause more hidden nodes and thus further cause performance degradation in the MAC protocol. However, lower received power levels will cause more hidden nodes and thus further cause performance degradation in the MAC protocol. Thus, when the problems of spectrum spatial-reuse efficiency and hidden nodes need to be solved at the same time, power management must be taken into account as an integral part of a MAC protocol. Thus, when the problems of spectrum spatial-reuse efficiency and hidden nodes need to be solved at the same time, power management must be taken into account as an integral part of a MAC protocol. 85

86. 86 Increase Transmission Rate Currently the physical layer of wireless networks usually support multiple transmission rates by using different modulation and coding schemes under different channel qualities. Currently the physical layer of wireless networks usually support multiple transmission rates by using different modulation and coding schemes under different channel qualities. The actual transmission rate is selected at the MAC layer according to the statistics of previous transmissions, and such a process is called MAC layer rate control. The actual transmission rate is selected at the MAC layer according to the statistics of previous transmissions, and such a process is called MAC layer rate control. 86

87. 87 Increase Transmission Rate In WMNs, multiple transmission paths exist. In WMNs, multiple transmission paths exist. As shown in Figure 7.2, considering end-to-end communications between Node A and Node C, the packets from Node A can go directly to Node C (called path 1), or go to node C via Node B (called path 2). As shown in Figure 7.2, considering end-to-end communications between Node A and Node C, the packets from Node A can go directly to Node C (called path 1), or go to node C via Node B (called path 2). 87

88. 88 Increase Transmission Rate By using path 1, the link quality between Node A and Node C may be poor and thus a low transmission rate has to be adopted. By using path 1, the link quality between Node A and Node C may be poor and thus a low transmission rate has to be adopted. If Node A increases its transmit power level, a higher transmission rate can be used. If Node A increases its transmit power level, a higher transmission rate can be used. However, Node A will generate a very large interference range. However, Node A will generate a very large interference range. On the other hand, if path 2 is adopted, both Node A and Node B can use a lower transmit power level to achieve a higher transmission rate. On the other hand, if path 2 is adopted, both Node A and Node B can use a lower transmit power level to achieve a higher transmission rate. 88

89. 89 Increase Transmission Rate The disadvantage in this scenario is that multi-hop transmission reduces capacity of the entire network, because one packet from Node A to Node C simply use the same channel for longer time. The disadvantage in this scenario is that multi-hop transmission reduces capacity of the entire network, because one packet from Node A to Node C simply use the same channel for longer time. Thus, a tradeoff is needed to determine if path 1 or path 2 is better. Thus, a tradeoff is needed to determine if path 1 or path 2 is better. The tradeoff needs to consider the power management, rate control and routing altogether. The tradeoff needs to consider the power management, rate control and routing altogether. 89

90. 90 Open Topics So far, the research of power management for mesh routers is mostly focused on reducing interference. So far, the research of power management for mesh routers is mostly focused on reducing interference. However, other important functionalities have not been investigated. However, other important functionalities have not been investigated. Thus, it is necessary to develop new power management schemes to self-organize network, increase transmission rate, and improve network connectivity and stability. Thus, it is necessary to develop new power management schemes to self-organize network, increase transmission rate, and improve network connectivity and stability. A more interesting research topic is to develop a power management scheme that support all functionalities discussed above. A more interesting research topic is to develop a power management scheme that support all functionalities discussed above. 90

91. 91 Power Management for Mesh Clients In contrast to mesh routers, mesh clients may expect protocols to be power efficient. In contrast to mesh routers, mesh clients may expect protocols to be power efficient. For example, some mesh clients are IP phones or even sensors; power efficiency is the major concern for them. For example, some mesh clients are IP phones or even sensors; power efficiency is the major concern for them. Thus, it is quite possible that some applications of WMNs require power management to optimize both power efficiency and connectivity, which results in a complicated problem. Thus, it is quite possible that some applications of WMNs require power management to optimize both power efficiency and connectivity, which results in a complicated problem. 91

92. 92 Power Management for Mesh Clients To avoid this complexity and also to take advantages of the availability of mesh routers, a rule of thumb for protocol design is to make sure protocols on mesh clients must be as simple as possible. To avoid this complexity and also to take advantages of the availability of mesh routers, a rule of thumb for protocol design is to make sure protocols on mesh clients must be as simple as possible. For example, for a routing protocol, the major computation and overhead must be kept inside mesh routers. For example, for a routing protocol, the major computation and overhead must be kept inside mesh routers. 92

93. 93 Open Research Issues To further improve the performance of WMNs, solutions are needed to the following research problems: To further improve the performance of WMNs, solutions are needed to the following research problems: –Developing a single power management scheme that can self-organize network, achieve high spectrum spatial- reuse efficiency, maintain stable network connectivity, and transmit packets using high transmission rate. 93

94. 94 Open Research Issues –Cross-layer design between power management and other protocols, in particular MAC and routing protocols. –Reducing protocol complexity in mesh clients. –Implementing protocols on mesh routers to support power saving mode for mesh clients. 94

95. 95 Topology Control and Management The topology of a wireless network depends on two sets of factors: uncontrollable factors and controllable factors The topology of a wireless network depends on two sets of factors: uncontrollable factors and controllable factors —Uncontrollable factors include noise, channel fading, etc. —Controllable factors can be transmit power level, antenna direction, and node locations. —Some factors cann be controllable or uncontrollable, depending on applications, e.g., node mobility, interference 95

96. 96 Topology Control and Management Topology control is an optimization process of utilizing controllable factors to compensate uncontrollable factors so that the network topology is formed into a structure that can achieve a desired performance level. Topology control is an optimization process of utilizing controllable factors to compensate uncontrollable factors so that the network topology is formed into a structure that can achieve a desired performance level. The performance metrics include energy efficiency, interference, stability, etc. The performance metrics include energy efficiency, interference, stability, etc. 96

97. 97 Topology Control and Management Based on different performance metrics considered in topology control, existing topology control schemes can be classified into the following major types Based on different performance metrics considered in topology control, existing topology control schemes can be classified into the following major types Topology control for energy efficiency: Topology control for energy efficiency: Most of existing topology control schemes are focused on the problem of reducing energy consumption Most of existing topology control schemes are focused on the problem of reducing energy consumption This is very meaningful for mobile ad hoc networks and wireless sensor networks, but not for WMNs. This is very meaningful for mobile ad hoc networks and wireless sensor networks, but not for WMNs. 97

98. 98 Topology Control and Management Some other schemes consider energy efficiency by Some other schemes consider energy efficiency by switching nodes into power save mode switching nodes into power save mode Thus, topology control is not only cross-related with Thus, topology control is not only cross-related with power management, but becomes more complicated as power management, but becomes more complicated as radio on/off directly changes the network topology. radio on/off directly changes the network topology. Again, this kind of methodology is not necessary for Again, this kind of methodology is not necessary for WMNs because of no or low constraints in energy. WMNs because of no or low constraints in energy. 98

99. 99 Topology Control for Low Interference Interference can be reduced by decreasing transmit power or lowering the network degree. Interference can be reduced by decreasing transmit power or lowering the network degree. However, to directly control the interference in a topology control scheme, the optimization problem must consider interference as a performance metric. However, to directly control the interference in a topology control scheme, the optimization problem must consider interference as a performance metric. 99

100. 100 Topology Control for Low Interference Reducing interference can increase transmission opportunity of an individual node, because its experienced interference becomes lower. Reducing interference can increase transmission opportunity of an individual node, because its experienced interference becomes lower. However, it does not guarantee that the network capacity also However, it does not guarantee that the network capacity also increases, since reducing interference is brought by low transmit increases, since reducing interference is brought by low transmit power, and thus cause a larger number of hops for the same power, and thus cause a larger number of hops for the same end-to-end communication. end-to-end communication. Thus, merely relying on interference to optimize topology Thus, merely relying on interference to optimize topology control may not be sufficient to improve overall network control may not be sufficient to improve overall network performance. performance. 100

101. 101 Topology Control for Fault Tolerance If topology control is only concerned with energy If topology control is only concerned with energy efficiency or reducing interference, the network topology efficiency or reducing interference, the network topology may becomes more sensitive to node failure or node may becomes more sensitive to node failure or node departure, for two reasons. departure, for two reasons. - First, the network may have few links available for - First, the network may have few links available for fault tolerance. fault tolerance. - Second, end-to-end communications between two - Second, end-to-end communications between two nodes have more hops, and thus become more fragile nodes have more hops, and thus become more fragile to node failure or node departure. to node failure or node departure. 101

102. 102 Topology Control Topology impacts the capacity of multi-hop networks. Topology impacts the capacity of multi-hop networks. However, no schemes have really consider capacity as a direct However, no schemes have really consider capacity as a direct performance metric, because network capacity is determined by performance metric, because network capacity is determined by several interleaved parameters such as transmit power level, several interleaved parameters such as transmit power level, interference, and average hop of a traffic flow, and also by interference, and average hop of a traffic flow, and also by several protocols. several protocols. 102

103. 103 Topology Control and Management Today some WMNs are still using single channel, while others have started to use multiple channels in order to increase network capacity. Today some WMNs are still using single channel, while others have started to use multiple channels in order to increase network capacity. Topology control for these two types of WMNs are different, since the channel allocation in a multi- channel WMN also impacts the topology. Topology control for these two types of WMNs are different, since the channel allocation in a multi- channel WMN also impacts the topology. 103

104. 104 Topology Control of Single-Channel Network So far few research work on topology control has been So far few research work on topology control has been specifically done for WMNs, even for single-channel specifically done for WMNs, even for single-channel WMNs. WMNs. The research work on mobile ad hoc networks may not be The research work on mobile ad hoc networks may not be applied to WMNs, since mobility is not a concern for applied to WMNs, since mobility is not a concern for topology control of WMNs. topology control of WMNs. Thus, node mobility must be taken out from the Thus, node mobility must be taken out from the optimization problem when a topology control algorithm is optimization problem when a topology control algorithm is designed for WMNs designed for WMNs 104

105. 105 Energy efficiency is not a constraint to the topology control of WMNs either. Energy efficiency is not a constraint to the topology control of WMNs either. Thus, the optimization problem does not need to consider the constraint of energy consumption. Thus, the optimization problem does not need to consider the constraint of energy consumption. In addition, the topology control scheme had no need to handle nodes in sleep or power save mode. In addition, the topology control scheme had no need to handle nodes in sleep or power save mode. This potentially makes the topology control problem in WMNs easier than that in mobile ad hoc networks, and thus, motivates us to develop a light-weight or practical topology control scheme for WMNs. This potentially makes the topology control problem in WMNs easier than that in mobile ad hoc networks, and thus, motivates us to develop a light-weight or practical topology control scheme for WMNs. 105 Topology Control of Single-Channel Network

106. 106 However, all of these schemes consider only one However, all of these schemes consider only one performance metric and take into account the relationship performance metric and take into account the relationship with only one protocol layer. with only one protocol layer. To best take advantages of topology control, a unified To best take advantages of topology control, a unified scheme that considers multiple performance metrics and scheme that considers multiple performance metrics and also multiple protocol layers is needed. also multiple protocol layers is needed. 106 Topology Control of Single-Channel Network

107. 107 Topology Control of Multi-Channel Network Multi-channel WMNs can be based on a single radio or multiple radios. Multi-channel WMNs can be based on a single radio or multiple radios. If it is single radio based, usually only one transceiver is available on the radio. If it is single radio based, usually only one transceiver is available on the radio. Thus, channels must be switched on time. Thus, channels must be switched on time. In other words, if not carefully designed in topology control, the network can be easily partitioned into different networks from time to time. In other words, if not carefully designed in topology control, the network can be easily partitioned into different networks from time to time. 107

108. 108 A simple approach to this problem is to use a common A simple approach to this problem is to use a common channel at a given time slot for all nodes. channel at a given time slot for all nodes. Thus, all nodes will have a consistent view on the network Thus, all nodes will have a consistent view on the network topology. topology. A topology control algorithm for a multichannel WMN is A topology control algorithm for a multichannel WMN is usually more complicated than that for a single-channel usually more complicated than that for a single-channel MAC, even if a single performance metric is considered. MAC, even if a single performance metric is considered. 108 Topology Control of Multi-Channel Network

109. 109 When multiple performance metrics are taken in account, the complexity is even higher. When multiple performance metrics are taken in account, the complexity is even higher. When a WMN consists of multi-radio nodes, the network partition issue can be easily resolved, because each node can use different radios at the same time to communicate with its neighbors. When a WMN consists of multi-radio nodes, the network partition issue can be easily resolved, because each node can use different radios at the same time to communicate with its neighbors. For the same reason, the optimization problem of topology control in a multi radio can be simpler; it does not need to consider channel sharing in a time division among nodes in the same neighbor. For the same reason, the optimization problem of topology control in a multi radio can be simpler; it does not need to consider channel sharing in a time division among nodes in the same neighbor. 109 Topology Control of Multi-Channel Network

110. 110 Open Research Issues: Topology Control with Directional Antenna So far most research work of topology control is focused on omni-directional antenna. So far most research work of topology control is focused on omni-directional antenna. Although directional antennas are difficult to use in mobile ad hoc networks, they have a good fit for WMNs, since minimal mobility in mesh nodes make changing antenna direction become a feasible task. Although directional antennas are difficult to use in mobile ad hoc networks, they have a good fit for WMNs, since minimal mobility in mesh nodes make changing antenna direction become a feasible task. 110

111. 111 Open Research Issues: Topology control with Directional Antenna Since different direction of antenna results in different link connectivity, so antenna direction control is closely coupled with topology control. Since different direction of antenna results in different link connectivity, so antenna direction control is closely coupled with topology control. Since both MAC and routing are impacted by antenna directions, topology control schemes for WMNs with directional antenna need to take into account the design of MAC protocols and routing protocols Since both MAC and routing are impacted by antenna directions, topology control schemes for WMNs with directional antenna need to take into account the design of MAC protocols and routing protocols 111

112. 112 Open Research Issues: Topology control for WMNs with heterogeneous nodes When nodes with single channel and nodes with different number of channels co-exist in the network, new topology control schemes need to be developed. When nodes with single channel and nodes with different number of channels co-exist in the network, new topology control schemes need to be developed. Moreover, nodes may also be different in maximum transmit power, which gives another constraint to the topology control scheme. Moreover, nodes may also be different in maximum transmit power, which gives another constraint to the topology control scheme. 112

113. 113 Open Research Issues: Cross-layer design with other protocols Since topology is actually used by routing either explicitly Since topology is actually used by routing either explicitly or implicitly, it is closely related to routing. or implicitly, it is closely related to routing. Moreover, topology also reflects the connectivity between Moreover, topology also reflects the connectivity between nodes in neighbor, so it also impacts the performance of nodes in neighbor, so it also impacts the performance of the MAC. the MAC. We also know that both routing protocol and MAC We also know that both routing protocol and MAC protocol are closely related to rate control. protocol are closely related to rate control. 113

114. 114 Open Research Issues: Cross-layer design with other protocols In addition, topology control is usually performed through transmit power adjustment, which further impacts power control and management schemes. In addition, topology control is usually performed through transmit power adjustment, which further impacts power control and management schemes. Therefore, the optimization of topology control must take into account all these constraints, which is a challenging research problem. Therefore, the optimization of topology control must take into account all these constraints, which is a challenging research problem. 114

115. 115 Open Research Issues: Topology control for stability Network stability is also impacted by variations in link quality, which is fast in time scale. Network stability is also impacted by variations in link quality, which is fast in time scale. Thus, new scheme is needed in topology control so that fast variable link quality will not make frequent change in topology. Thus, new scheme is needed in topology control so that fast variable link quality will not make frequent change in topology. Otherwise, neither MAC nor routing will work well with satisfactory performance. Otherwise, neither MAC nor routing will work well with satisfactory performance. 115

116. 116 Open Research Issues: Low complexity and practical topology control solutions To date, many topology control schemes are proposed without considering the computation complexity or possibility of being implemented. To date, many topology control schemes are proposed without considering the computation complexity or possibility of being implemented. Although the performance is justified through theoretical analysis, they are difficult to be implemented. Although the performance is justified through theoretical analysis, they are difficult to be implemented. 116

117. 117 Traditional Network Management Functions for WMNs In addition to the above-mentioned functions that are new to WMNs, the traditional network management functions are also indispensable components of network management for WMNs. In addition to the above-mentioned functions that are new to WMNs, the traditional network management functions are also indispensable components of network management for WMNs. However, these functions need to be enhanced, since they are initially developed for wired networks and have not considered the specific features of WMNs. However, these functions need to be enhanced, since they are initially developed for wired networks and have not considered the specific features of WMNs. 117

118. 118 Traditional Network Management Functions for WMNs Considering the example of network monitoring and fault tolerance management, there are many new challenges in WMNs. Considering the example of network monitoring and fault tolerance management, there are many new challenges in WMNs. Firstly, the statistics in the management information base (MIB) of mesh nodes have to traverse multiple hops to arrive at a sever. Firstly, the statistics in the management information base (MIB) of mesh nodes have to traverse multiple hops to arrive at a sever. 118

119. 119 Traditional Network Management Functions for WMNs Secondly, data processing algorithms for fault detection become more complicated because much more possibilities of faults exist in a multi-hop wireless networks. Secondly, data processing algorithms for fault detection become more complicated because much more possibilities of faults exist in a multi-hop wireless networks. Finally, when fault is detected, the action to be taken is also more difficult to complete: notification is difficult to send, exact point of errors is hard to identify, and also automatic fix becomes a challenge. Finally, when fault is detected, the action to be taken is also more difficult to complete: notification is difficult to send, exact point of errors is hard to identify, and also automatic fix becomes a challenge. 119

120. 120 Traditional Network Management Functions In addition, in order to accurately detect abnormal operation of In addition, in order to accurately detect abnormal operation of WMNs, effective data processing algorithms are needed. WMNs, effective data processing algorithms are needed. Overall, research efforts and development practice are needed Overall, research efforts and development practice are needed to investigate how to incorporate the existing network to investigate how to incorporate the existing network management functions into the architecture of WMNs. management functions into the architecture of WMNs. 120