27-08-2014, 02:59 PM
4G TEHNOLOGY-A SENSATION ON SEMINAR REPORT
[attachment=67165]
Abstract
This paper describes an architecture for differentiation of Quality of Service in heterogeneous wireless-wired networks. This architecture applies an “all-IP” paradigm, with embedded mobility of users. The architecture allows for multiple types of access networks, and enables user roaming between different operator domains. The overall 4G architecture discussed in this paper is IPv6-based, supporting seamless mobility between different access technologies. Mobility is a substantial problem in such environment, because inter-technology handovers have to be supported. In our case, we targeted Ethernet (802.3) for wired access; Wi-Fi (802.11b) for wireless LAN access; and W-CDMA - the radio interface of UMTS - for cellular access.The architecture is able to provide quality of service per-user and per-service An integrated service and resource management approach is presented based on the cooperative association between Quality
of Service Brokers and Authentication, Authorisation, Accounting and Charging systems. The different phases of QoS-operation are discussed. The overall QoS concepts are presented with some relevant enhancements that address specifically voice services. In particular, EF simulations results are discussed in this context
INTRODUCTION
A wireless network is an infrastructure for communication “through the air”, in other words, no cables are needed to connect from one point to another. These connections can be used for speech, e-mail, surfing on the Web and transmission of audio and video. The most widespread use is mobile telephones. Wireless networks are also used for communication between computers. This note focuses on ways to set up wireless connections between computers. It gives a basic overview without becoming too technical. It will help to determine whether a wireless network might be a suitable solution. It also is a guide to more resources. Many links are to a document by Mike Jensen. The links used are examples; they are not preferred products
GENERATIONS OF WIRELESS COMMUNICATION
These first generation mobile systems were designed to offer a single service that is speech.
2G: These second generation mobile systems were also designed primarily to offer speech with a limited capability to offer data at low rates.
3G: These third generation mobile systems are expected to offer high quality multimedia services and operative different environments. These systems are referred to as universal mobile telecommunication systems (UMTS) in Europe and international mobile telecommunication systems 2000(IMT2000) worldwide.
4G: This is user-driven, user controlled services and context aware applications. Compared to 3G ,4G has higher data rates and it has QOS which is the main criteria in 4G wireless commuication.
Availability of the network services anywhere, at anytime, can be one of the key factors that attract individuals and institutions to the new network infrastructures, stimulate the development of telecommunications, and propel economies. This bold idea has already made its way into the telecommunication community bringing new requirements for network design, and envisioning a change of the current model of providing services to customers. The emerging new communications paradigm assumes a user to be able to access services independently of her or his location, in an almost transparent way, with the terminal being able to pick the preferred access technology at current location (ad-hoc, wired, wireless LAN, or cellular), and move between technologies seamlessly i.e. without noticeable disruption. Unified, secure, multi-service, and multiple-operator network architectures are now being developed in a context commonly referenced to as networks Beyond-3G or, alternatively, 4G networks
AN ALL-IP 4G NETWORK ARCHITECTURE
The overall 4G architecture discussed in this paper is IPv6-based, supporting seamless mobility between different access technologies. Mobility is a substantial problem in such environment, because inter-technology handovers have to be supported. In our case, we targeted Ethernet (802.3) for wired access; Wi-Fi (802.11b) for wireless LAN access; and W-CDMA - the radio interface of UMTS - for cellular access (Fig. 1). With this diversity, mobility cannot be simply handled by the lower layers, but needs to be implemented at the network layer. An "IPv6-based" mechanism has to be used for interworking, and no technology-internal mechanisms for handover, neither on the wireless LAN nor on other technology, can be used. So, in fact no mobility mechanisms are supported in the W-CDMA cells, but instead the same IP protocol supports the movement between cells. Similarly, the 802.11 nodes are only in BSS modes, and will not create an ESS: IPv6 mobility will handle handover between cells. 1 The concepts that are presented in this paper have been developed and tested in controlled environments in the IST project Moby Dick [2] and are currently being refined
PROVIDING QUALITY OF SERVICE
The design principle for QoS architecture was to have a structure which allows for a potentially scalable system that can maintain contracted levels of QoS. Eventually, especially if able to provide an equivalent to the Universal Telephone Service, it could possibly replace today's telecommunications networks. Therefore, no specific network services should be presumed nor precluded, though the architecture should be optimised for a representative set of network services. Also, no special charging models should be imposed by the AAAC system, and the overall architecture must be able to support very restrictive network resource usage. In terms of services, applications that use VoIP, video streaming, web, e-mail access and file transfer have completely different prerequisites, and the network should be able to differentiate their service. The scalability concerns favour a differentiated services (DiffServ) approach [5]. This approach is laid on theassumption to control the requests at the borders of the network, and that end-to-end QoS assurance is achieved by a concatenation of multiple managed entities. With such requirements, network resource control must be under the control of the network service provider. It has to be able to control every resource, and to grant or deny user and service access. This requirement calls for flexible and robust explicit connections admission control (CAC) mechanisms at the network edge, able to take fast decisionson user requests.
Service and Network Management in Mobile Networks
Our approach for 4G networks and to service provisioning is based on the separation of service and network management entities. In our proposal we define a service layer, which has its own interoperation mechanisms across different administrative domains (and can be mapped to the service provider concept), and a network layer, which has its own interoperation mechanism between network domains. An administrative domain may be composed of one or more technology domains. Service definitions are handled inside administrative domains and service translation is done between administrative domains [6]. Each domain has an entity responsible for handling user service aspects (the AAAC system), and at least one entity handling the network resource management aspects at the access level (the QoS Broker). The AAAC system is the central point for Authentication, Authorization and Accounting. When a mobile user enters the network, the AAAC is supposed to authenticate him. Upon successful authentication, the AAAC sends to the QoS Broker the relevant QoS policy information based on the SLA of the user, derived from his profile. From then, it is assumed that the AAAC has delegated
Implicit "Session" Signalling
In this architecture, each network service being offered in the network is associated to a different DSCP code. This way, every packet has the information needed to the network entities to correctly forward, account, and differentiate service delivered to different packets. After registering (with the AAAC system) a user application can “signal” the intention of using a service by sending packets marked with appropriate DSCP. These packets are sent in a regular way in wired access networks, or over a shared uplink channel used for signalling in W-CDMA. This way of requesting services corresponds to implicit signalling, user-dependent, as the QoS Broker will be aware of the semantics of each DSCP code per each user (although typically there will be no variation on the meaning of DSCP codes between users). Thus QoS Broker has the relevant information for mapping user-service requests into network resources requirements and based on this information configures an access router.A novel concept of “session” is implemented: the concept of a “session” is here associated with the usage of specific network resources, and not explicitly with specific traffic micro-flows. This process is further detailed in section 4.
Network services offer
Services will be ofered a the network operator independently on the user applications, but will be flexible enough to support user applications Each offered network service will be implemented with one of the three basic DiffServ per-hop behaviours (EF, AF, or BE), with associated bandwidth characteristics. Table 1 lists the network services used in the tests. The network services include support for voice communications (e.g. via S1) and data transfer services. Delay, delay jitter and packet loss rate are among the possible parameters to include in the future, but no specific control mechanisms for these parameters are currently used. The services may also be unidirectional or bi-directional. In fact, the QoS architecture can support any type of network service, where the only limit is the level of management complexity expressed in terms of complexity of interaction between the QoS Brokers, the AAAC systems and the AR that the network provider is willing to support.
Registration and Authorisation
The Registration process (Figure 2) is initiated after a Care of Address (CoA) is acquired by the MT via stateless auto-configuration, avoiding Duplicate Address Detection (DAD) by using unique layer-2 identifiers [7] to create the Interface Identifier part of the IPv6 address. However, getting a CoA does not entitle the user to use resources, besides registration messages and emergency calls. The MT has to start the authentication process by exchanging the authentication information with the AAAC through the AR. Upon a successful authentication, the AAAC System will push the NVUP (network view of the User Profile) to both the QoS Broker and the MT, via the AR. Messages 1 to 4 on Figure 2 detail this process. The same picture shows how each network service is authorized (messages 5 to 8). The packets sent from the MT with a specific DSCP implicit signal the request of a particular service, such as a voice call (supported by network service S1, as in Table 1). If the requested service does not match any policy already set in the AR (that is, the user has not established a voice call before, e.g.), the QoS attendant/manager at the AR interacts with the QoS Broker that analyses the request and authorises the service or not, based on the User NVUP (Network View of the User Profile) and on the availability of resources. This authorisation corresponds to a configuration of the AR (via COPS [10]) with the appropriate policy for that user and that service (e.g. allowing the packets marked as “belonging” to voice call to go through, and
configuring the proper scheduler parameters, as we will see in section 4.3). After that, packets with authorised profile will be let into the network and non-conformant packets will restart the authorization process once more, or will be discarded
EF PHB resource provisioning
Building an all-IP architecture based on a Differentiated Services introduces a problem of how to create per-domain services for transport of traffic aggregates with a given QoS. Per-domain services support data exchange by mixing traffic of different applications, therefore different aggregates are required to support delay-sensitive traffic, delay tolerant traffic, inelastic, elastic, as well as network maintenance traffic (e.g. SNMP, DNS, COPS, AAAC etc.). As applications generate traffic of different characteristics in terms of data rates, level of burstiness, packet size distribution and because the operator needs to protect the infrastructure against congestion, it is very important that aggregate scheduling will be accompanied by:
per-user rate limitation performed in the ingress routers (ARs) based on user profile,
dimensioning and configuration of network resources to allow for a wide range of user needs and services,
resource management for edge-to-edge QoS.
CONCLUSION
We presented an architecture for supporting end-to-end QoS. This QoS architecture is able to support multi-service, multi-operator environments, handling complex multimedia services, with per user and per service differentiation, and integrating mobility and AAAC aspects. The main elements in our architecture are the MT, the AR and the QoS Brokers. We discussed the simple interoperation between these elements and depicted the overall QoS concept. With our approach, very little restrictions are imposed on the service offering. This architecture is currently being evolved for large testing in field trials across Madrid and Stuttgart. Being an architecture specially targeted to support real time communications over packet networks, the network elements configuration must be well dissected. The simulation study summarized in the paper was a valuable input to the QoS Broker implementation and policies design, providing simple heuristics to properly configure the access routers to achieve the best possible performance. The schedulers configuration on the core routers was also determined through the results of this simulation study. This architecture still has some shortcomings, though, mostly due to its diffserv orientation. Each domain has to implement its own plan for mapping between network service and a DSCP, and thus, for inter domain service provision, it is essential a service/DSCP mapping between neighbouring domains. Furthermore, an adequate middleware function is required in the MT, to optimally mark the packets generated by the applications and issue the proper service requests, which requires extensions in current protocol stacks
[attachment=67165]
Abstract
This paper describes an architecture for differentiation of Quality of Service in heterogeneous wireless-wired networks. This architecture applies an “all-IP” paradigm, with embedded mobility of users. The architecture allows for multiple types of access networks, and enables user roaming between different operator domains. The overall 4G architecture discussed in this paper is IPv6-based, supporting seamless mobility between different access technologies. Mobility is a substantial problem in such environment, because inter-technology handovers have to be supported. In our case, we targeted Ethernet (802.3) for wired access; Wi-Fi (802.11b) for wireless LAN access; and W-CDMA - the radio interface of UMTS - for cellular access.The architecture is able to provide quality of service per-user and per-service An integrated service and resource management approach is presented based on the cooperative association between Quality
of Service Brokers and Authentication, Authorisation, Accounting and Charging systems. The different phases of QoS-operation are discussed. The overall QoS concepts are presented with some relevant enhancements that address specifically voice services. In particular, EF simulations results are discussed in this context
INTRODUCTION
A wireless network is an infrastructure for communication “through the air”, in other words, no cables are needed to connect from one point to another. These connections can be used for speech, e-mail, surfing on the Web and transmission of audio and video. The most widespread use is mobile telephones. Wireless networks are also used for communication between computers. This note focuses on ways to set up wireless connections between computers. It gives a basic overview without becoming too technical. It will help to determine whether a wireless network might be a suitable solution. It also is a guide to more resources. Many links are to a document by Mike Jensen. The links used are examples; they are not preferred products
GENERATIONS OF WIRELESS COMMUNICATION
These first generation mobile systems were designed to offer a single service that is speech.
2G: These second generation mobile systems were also designed primarily to offer speech with a limited capability to offer data at low rates.
3G: These third generation mobile systems are expected to offer high quality multimedia services and operative different environments. These systems are referred to as universal mobile telecommunication systems (UMTS) in Europe and international mobile telecommunication systems 2000(IMT2000) worldwide.
4G: This is user-driven, user controlled services and context aware applications. Compared to 3G ,4G has higher data rates and it has QOS which is the main criteria in 4G wireless commuication.
Availability of the network services anywhere, at anytime, can be one of the key factors that attract individuals and institutions to the new network infrastructures, stimulate the development of telecommunications, and propel economies. This bold idea has already made its way into the telecommunication community bringing new requirements for network design, and envisioning a change of the current model of providing services to customers. The emerging new communications paradigm assumes a user to be able to access services independently of her or his location, in an almost transparent way, with the terminal being able to pick the preferred access technology at current location (ad-hoc, wired, wireless LAN, or cellular), and move between technologies seamlessly i.e. without noticeable disruption. Unified, secure, multi-service, and multiple-operator network architectures are now being developed in a context commonly referenced to as networks Beyond-3G or, alternatively, 4G networks
AN ALL-IP 4G NETWORK ARCHITECTURE
The overall 4G architecture discussed in this paper is IPv6-based, supporting seamless mobility between different access technologies. Mobility is a substantial problem in such environment, because inter-technology handovers have to be supported. In our case, we targeted Ethernet (802.3) for wired access; Wi-Fi (802.11b) for wireless LAN access; and W-CDMA - the radio interface of UMTS - for cellular access (Fig. 1). With this diversity, mobility cannot be simply handled by the lower layers, but needs to be implemented at the network layer. An "IPv6-based" mechanism has to be used for interworking, and no technology-internal mechanisms for handover, neither on the wireless LAN nor on other technology, can be used. So, in fact no mobility mechanisms are supported in the W-CDMA cells, but instead the same IP protocol supports the movement between cells. Similarly, the 802.11 nodes are only in BSS modes, and will not create an ESS: IPv6 mobility will handle handover between cells. 1 The concepts that are presented in this paper have been developed and tested in controlled environments in the IST project Moby Dick [2] and are currently being refined
PROVIDING QUALITY OF SERVICE
The design principle for QoS architecture was to have a structure which allows for a potentially scalable system that can maintain contracted levels of QoS. Eventually, especially if able to provide an equivalent to the Universal Telephone Service, it could possibly replace today's telecommunications networks. Therefore, no specific network services should be presumed nor precluded, though the architecture should be optimised for a representative set of network services. Also, no special charging models should be imposed by the AAAC system, and the overall architecture must be able to support very restrictive network resource usage. In terms of services, applications that use VoIP, video streaming, web, e-mail access and file transfer have completely different prerequisites, and the network should be able to differentiate their service. The scalability concerns favour a differentiated services (DiffServ) approach [5]. This approach is laid on theassumption to control the requests at the borders of the network, and that end-to-end QoS assurance is achieved by a concatenation of multiple managed entities. With such requirements, network resource control must be under the control of the network service provider. It has to be able to control every resource, and to grant or deny user and service access. This requirement calls for flexible and robust explicit connections admission control (CAC) mechanisms at the network edge, able to take fast decisionson user requests.
Service and Network Management in Mobile Networks
Our approach for 4G networks and to service provisioning is based on the separation of service and network management entities. In our proposal we define a service layer, which has its own interoperation mechanisms across different administrative domains (and can be mapped to the service provider concept), and a network layer, which has its own interoperation mechanism between network domains. An administrative domain may be composed of one or more technology domains. Service definitions are handled inside administrative domains and service translation is done between administrative domains [6]. Each domain has an entity responsible for handling user service aspects (the AAAC system), and at least one entity handling the network resource management aspects at the access level (the QoS Broker). The AAAC system is the central point for Authentication, Authorization and Accounting. When a mobile user enters the network, the AAAC is supposed to authenticate him. Upon successful authentication, the AAAC sends to the QoS Broker the relevant QoS policy information based on the SLA of the user, derived from his profile. From then, it is assumed that the AAAC has delegated
Implicit "Session" Signalling
In this architecture, each network service being offered in the network is associated to a different DSCP code. This way, every packet has the information needed to the network entities to correctly forward, account, and differentiate service delivered to different packets. After registering (with the AAAC system) a user application can “signal” the intention of using a service by sending packets marked with appropriate DSCP. These packets are sent in a regular way in wired access networks, or over a shared uplink channel used for signalling in W-CDMA. This way of requesting services corresponds to implicit signalling, user-dependent, as the QoS Broker will be aware of the semantics of each DSCP code per each user (although typically there will be no variation on the meaning of DSCP codes between users). Thus QoS Broker has the relevant information for mapping user-service requests into network resources requirements and based on this information configures an access router.A novel concept of “session” is implemented: the concept of a “session” is here associated with the usage of specific network resources, and not explicitly with specific traffic micro-flows. This process is further detailed in section 4.
Network services offer
Services will be ofered a the network operator independently on the user applications, but will be flexible enough to support user applications Each offered network service will be implemented with one of the three basic DiffServ per-hop behaviours (EF, AF, or BE), with associated bandwidth characteristics. Table 1 lists the network services used in the tests. The network services include support for voice communications (e.g. via S1) and data transfer services. Delay, delay jitter and packet loss rate are among the possible parameters to include in the future, but no specific control mechanisms for these parameters are currently used. The services may also be unidirectional or bi-directional. In fact, the QoS architecture can support any type of network service, where the only limit is the level of management complexity expressed in terms of complexity of interaction between the QoS Brokers, the AAAC systems and the AR that the network provider is willing to support.
Registration and Authorisation
The Registration process (Figure 2) is initiated after a Care of Address (CoA) is acquired by the MT via stateless auto-configuration, avoiding Duplicate Address Detection (DAD) by using unique layer-2 identifiers [7] to create the Interface Identifier part of the IPv6 address. However, getting a CoA does not entitle the user to use resources, besides registration messages and emergency calls. The MT has to start the authentication process by exchanging the authentication information with the AAAC through the AR. Upon a successful authentication, the AAAC System will push the NVUP (network view of the User Profile) to both the QoS Broker and the MT, via the AR. Messages 1 to 4 on Figure 2 detail this process. The same picture shows how each network service is authorized (messages 5 to 8). The packets sent from the MT with a specific DSCP implicit signal the request of a particular service, such as a voice call (supported by network service S1, as in Table 1). If the requested service does not match any policy already set in the AR (that is, the user has not established a voice call before, e.g.), the QoS attendant/manager at the AR interacts with the QoS Broker that analyses the request and authorises the service or not, based on the User NVUP (Network View of the User Profile) and on the availability of resources. This authorisation corresponds to a configuration of the AR (via COPS [10]) with the appropriate policy for that user and that service (e.g. allowing the packets marked as “belonging” to voice call to go through, and
configuring the proper scheduler parameters, as we will see in section 4.3). After that, packets with authorised profile will be let into the network and non-conformant packets will restart the authorization process once more, or will be discarded
EF PHB resource provisioning
Building an all-IP architecture based on a Differentiated Services introduces a problem of how to create per-domain services for transport of traffic aggregates with a given QoS. Per-domain services support data exchange by mixing traffic of different applications, therefore different aggregates are required to support delay-sensitive traffic, delay tolerant traffic, inelastic, elastic, as well as network maintenance traffic (e.g. SNMP, DNS, COPS, AAAC etc.). As applications generate traffic of different characteristics in terms of data rates, level of burstiness, packet size distribution and because the operator needs to protect the infrastructure against congestion, it is very important that aggregate scheduling will be accompanied by:
per-user rate limitation performed in the ingress routers (ARs) based on user profile,
dimensioning and configuration of network resources to allow for a wide range of user needs and services,
resource management for edge-to-edge QoS.
CONCLUSION
We presented an architecture for supporting end-to-end QoS. This QoS architecture is able to support multi-service, multi-operator environments, handling complex multimedia services, with per user and per service differentiation, and integrating mobility and AAAC aspects. The main elements in our architecture are the MT, the AR and the QoS Brokers. We discussed the simple interoperation between these elements and depicted the overall QoS concept. With our approach, very little restrictions are imposed on the service offering. This architecture is currently being evolved for large testing in field trials across Madrid and Stuttgart. Being an architecture specially targeted to support real time communications over packet networks, the network elements configuration must be well dissected. The simulation study summarized in the paper was a valuable input to the QoS Broker implementation and policies design, providing simple heuristics to properly configure the access routers to achieve the best possible performance. The schedulers configuration on the core routers was also determined through the results of this simulation study. This architecture still has some shortcomings, though, mostly due to its diffserv orientation. Each domain has to implement its own plan for mapping between network service and a DSCP, and thus, for inter domain service provision, it is essential a service/DSCP mapping between neighbouring domains. Furthermore, an adequate middleware function is required in the MT, to optimally mark the packets generated by the applications and issue the proper service requests, which requires extensions in current protocol stacks