AAL2 Signalling Framework to Support a Gigabit AAL2 Switching Node

Transcription

1 Framework to Support a Gigabit Switching Node André van Zyl, Neco Ventura Department of Electrical Engineering, University of Cape Town, Rondebosch, Cape Town, South Africa {andre, neco}@crg.ee.uct.ac.za Keywords: ATM,, SIGNALLING, WUGS Abstract This paper examines an signalling protocol in a hybrid ATM/ network. After an overview of signalling is given, the important issues relating to signalling are discussed. These issues include the independence between ATM signalling and signalling, the choice of which AAL type to utilise in transporting the signalling messages as well as the importance of limiting the delays incurred in signalling. An signalling framework, which we intend to implement, is proposed and various design decisions based on the above mentioned issues are discussed. A test bed implementation is finally proposed to demonstrate the functionality of the signalling framework in an switching network. 1. INTRODUCTION Although Asynchronous Transfer Mode (ATM) Adaptation Layer type 2 () has only been in existence for a few short years, it is widely being adopted as the technology of choice in VoDSL [1], VTOA trunking [2], as well as in UMTS wireless networks [3]. Amongst the transportation of various network traffic types, real-time voice is the most challenging. The two most important concerns when transporting voice in a packet based network, are end-to-end delay and the efficient use of bandwidth. The maximum acceptable end-to-end delay for the transportation of voice, without the use of echo cancellation, is 25 ms. When echo is adequately controlled, end-to-end delay of up to 150 ms is acceptable [4]. The total end-to-end voice delay is calculated as the sum of the speech encoding delay, the packetisation delay and the transmission delay. Of these three delay sources, the packetisation delay is the most critical one in a packet based network, as it can result in the longest delay. Voice has traditionally been transported over ATM as uncompressed 64 Kb/s Circuit Emulation Service (CES), utilising AAL1, but this method has its limitations. AAL1 is not bandwidth efficient, since it was intended for Constant Bit Rate (CBR) traffic, but voice is Variable Bit Rate (VBR) by nature. The authors would like to thank Telkom SA, Siemens, National Research Foundation (NRF) and The Department of Trade and Industry (DTI) for supporting this research project. AAL1 also suffers from unacceptable packetisation delay when applied to low-bit-rate, compressed, voice traffic. Significant advances in voice compression technology in the last decade have necessitated the need for a new AAL type. In 1997 the ITU-T addressed this need and standardised AAL type 2 (I.363.2) to provide bandwidth efficient transmission of low-bit-rate, short and variable length packets in delay sensitive applications [5]. has since been used extensively for Loop Emulation Service (LES), by trunking compressed voice as a cost effective alternative to AAL1. transports voice in the form of small, variable length packets, called CPS packets. Each packet is fitted with a header that contains a Channel Identifier (CID). Multiple CPS packets, from various sources, are then multiplexed into a single ATM cell to reduce the packetisation delay. Although this method utilises the local loop, between the Customer Premises Equipment (CPE) and the ATM network, very efficiently it also introduces a drawback when not all the CPS packets are intended for the same destination. This problem can be addressed by introducing an additional switching level on top of ATM, called switching. Thus switching of CPS packets will be performed based upon the CID value of the individual CPS packet [6]. This switching technique utilises the core network resources more efficiently. The device capable of achieving this switching functionality is called an switching node. Research commenced into the design of an switching node at the University of Cape Town in the year An experimental ATM research switch, called the Washington University Gigabit Switch (WUGS) is being used to implement this switching node design. The design comprises of two distinct components, namely an user-plane switching component and an control-plane signalling component. The user-plane switching component, which forms part of another researcher s work, performs the underlying switching of individual CPS packets [7]. The control-plane signalling component provides the necessary signalling functionality required to establish end-to-end connections. ATM is a virtual circuit based technology and requires the establishment of an end-to-end connection prior to the transfer of user traffic. As a result, prior to two users transmitting traffic, an connection first needs to be established. This process of connection setup is done using

2 the procedures of signalling. signalling was recently standardised and provides the ability to dynamically establish, manage and release connections over existing ATM Virtual Circuits (VCs). This standard is often viewed as an extension to normal ATM signalling, but this is not the case. In this paper we present an signalling protocol framework, based on various available standards, which will reside in the control-plane of the switching node and on various end-points. The signalling protocol framework would need to dynamically establish new channels, end-to-end, without negatively impacting on existing channels, maintain those channels and tear them down. The combination of the signalling and switching components would result in the development of an switching node, which in practise would form part of a hybrid ATM / network. The rest of the paper is organized as follows: The following section will give a brief overview of signalling and its specification. Section 3 will consider various design issues that need to be considered relating to an signalling framework. Section 4 will look at the proposed test bed implementation and discuss its relevance. 2. OVERVIEW OF SIGNALLING The main objective of signalling is to establish, maintain and tear down connections between peer nodes. An node can be either an endpoint or an switch. is the only AAL layer that has its own signalling protocol. The primary reasons for this are as follows: Firstly, ATM has many signalling protocols, like the ATM Forum s UNI 4.0 and PNNI 1.0, the ITU-T s Q.2931, Q.2971 and B-ISUP. Extending all these protocols to accommodate would be very difficult. Secondly, having two separate signalling protocols has its advantages. The decoupling of the two signalling protocols would allow multiple overlay networks to operate over a single ATM network. Each network would operate with its own routing and addressing schemes [8]. This separation becomes crucial when different operators own and manage the two different networks. Finally, has the ability to multiplex many channels into a single ATM virtual connection. In order to establish and release these channels, a dynamic protocol is required, and this is accomplished by means of signalling. signalling messages can be carried as Channel Associated (CAS) or as Common Channel (CCS), also known as In-Band or Out-of-Band signalling. This issue, and the impact thereof, will be explained later in this paper. The signalling protocol is a protocol completely independent of ATM signalling. It assumes the existence of an ATM virtual channel between two users before channels are established inside the ATM VC. The ATM VC could be a PVC, SPVC or SVC. How these ATM virtual channels are established is not within the scope of the signalling protocol specification. The diagram below illustrates the signalling reference model. End-point Served Sig. Stack Underlying (e.g. MTP3b + NNI-SAAL, UNI-SAAL, SSCOP,...) Switch Sig. Stack Fig.1 Reference Model End-point Served Sig. Stack Underlying (e.g. MTP3b + NNI-SAAL, UNI-SAAL, SSCOP,...) The diagram contains three nodes, two endpoints and one switch. An end-point provides services like connection establishment and release and the switch provides switching and routing support. The (STC), situated below the Stack, provides the independence between the signalling protocol and the underlying signalling transport. The type of STC used will depend on the underlying signalling transport mechanism being used. Currently there are two STC standards. The ITU-T s Q is used if the transport mechanism is broadband MTP3b and Q is used if the transport mechanism is UNI-SAAL. Services provided by the STC will include assured (error free) transfer of data and in sequence delivery of data [9]. The most important issue still unresolved in the signalling protocol is the acquisition and distribution of routing information. The result is that most links are static and thus signalling is limited to small domains. In order for a completely independent overlay network to operate over an ATM network, an routing protocol is essential. The routing protocol s primary function would be to update the addressing / routing tables in the nodes in accordance to the network topology of the network. 3. DESIGN CONSIDERATIONS FOR THE SIGNALLING FRAMEWORK In designing the signalling framework certain issues relating to signalling need to be identified and addressed. The following sub-sections will look at some of the most important issues, but first the relationship between signalling and ATM signalling is explored. 3.1 RELATIONSHIP BETWEEN SIGNALLING AND ATM SIGNALLING As stated previously, these two signalling protocols are independent from each other, but signalling is unable to establish an end-to-end connection, between two nodes, without the existence of a previously established ATM VC between them. This ATM VC is established via ATM signalling. It is therefore not possible for an user to request a new connection, to another user, if these two users are without a prior established, underlying ATM VC. The ATM signalling protocol is required to establish such an

3 ATM virtual connection by means of a SETUP message. Some of the basic parameters contained within the ATM signalling SETUP message are, a) the ATM Destination Address, b) the Broadband Bearer Capacity, which contains the traffic type, e.g. CBR, VBR, etc., c) the Traffic Descriptor, d) the QoS parameters and e) the AAL parameters. Once an ATM VC is established, the signalling protocol will establish additional connections over this ATM connection, as requested by users. This is done using the signalling Establish Request (ERQ) message. As parameters the ERQ message contains, a) the Destination Address, b) the Connection Element Identifier (CEID), which includes the path identifier and the channel identifier (CID) and c) the Service specific information to define the type of SSCS residing over [10]. The network operator therefore needs to know beforehand the typical QoS and bandwidth requirements needed by an user. This knowledge would enable the operator to request the required QoS and traffic parameters from the ATM operator who will need to provide the necessary ATM VC. An example of a practical network topology will be illustrated and explained in section 4 to clarify the above statement. The basic exchange of signalling messages, between peer entities, is referred to as handshaking. The figure below illustrates the procedures required for call establishment and release in signalling. End-point Switch End-point Computer ERQ ECF REL RLC ERQ ECF REL RLC Computer ERQ - Establish Request ECF - Establish Confirm REL - Release Request RLC - Release Confirm Fig.2 Call establishment and release procedures in 3.2 CHOICE OF SIGNALLING TRANSPORTER One of the issues to be considered with signalling is which AAL layer to use when transporting the signalling messages. signalling messages can be carried as Channel Associated (CAS) or as Common Channel (CCS) as indicated in the diagram below. SIGNALLING STC (Q ) UNI - SSCF (Q.2130) SSCOP (Q.2110) SSSAR / SSTED (I.366.1) (I.363.2) ATM (I.361) PHYSICAL (I.432) over SIGNALLING STC (Q ) UNI - SSCF (Q.2130) SSCOP (Q.2110) AAL5 CPCS (I.363.5) AAL5 SAR (I.363.5) ATM (I.361) PHYSICAL (I.432) over AAL5 Fig.3 carried over or AAL5 When transported over an channel, CAS is used because the bearer traffic is carried in the same VC as the signalling messages. When using AAL5 as the transport layer, the signalling messages are carried in a dedicated ATM VC as CCS. The logical questions are thus: a) Should or AAL5 be used to transport signalling messages? b) What are the implications if the signalling messages and the bearer traffic are carried in the same VC? These questions were investigated by means of simulation [11] and it was found that signalling messages transported over has less signalling delay and thus a higher throughput than when using AAL5. With regard to the second question, simulation also showed that when signalling messages were sent in the same ATM VC as the bearer traffic, it was more beneficial to use two separate queues inside the node, instead of a single FIFO queue for both traffic types. One queue is dedicated to signalling messages and the other to normal traffic. The signalling queue would then also be assigned a higher priority than the traffic queue. The traffic queue will only be serviced once the signalling queue is empty. 3.3 SIGNALLING DELAY has been chosen as the transport technology for the access network in third generation UMTS wireless networks. The access network is required to support the Soft Handover (SHO) functionality, which requires a fast connection setup and teardown. This can be accomplished by minimising the signalling delay. The previous sub-section has already discussed two techniques that will contribute to the lowering of the signalling delay. It has also been proven, via simulation, that the signalling delay can be even further reduced when prioritised handling of signalling messages is introduced [12]. The signalling messages that should receive higher priority are Establish Request (ERQ) and Establish Confirm (ECF). These two messages are responsible for creating an connection. Two separate signalling

4 queues were used in the simulation. The first queue collects the ERQ and ECF signalling messages and only once this queue was empty, was the second queue serviced. This resulted in a 12% faster connection establishment time than using a normal FIFO queue for all signalling messages. Although the above two techniques could be used to lower the signalling delay, it has been shown via simulation [8] that the most beneficial technique is prioritising the ATM VCs carrying voice above the ATM VCs carrying data. The different techniques to be employed to lower the signalling delay can thus be summarised and listed in a hierarchical level of effectiveness as, a) voice traffic receives a higher priority than data traffic, b) signalling data receives a higher priority than voice traffic, and c) Establish Request (ERQ) and Establish Confirm (ECF) messages receives a higher priority than any other signalling messages. 3.4 HIGH-LEVEL SIGNALLING PROTOCOL ARCHITECTURE In this sub-section a high-level architecture of the signalling protocol framework with its various components will be illustrated and discussed. It depicts the Stack, as illustrated in figure 1 in more detail. The greyed boxes, in Figure 4, are those that we plan to develop and provide to the user. Functional Interface State Machine Initialisation Functional Interface Served Interface Functions Control Module Interface Functions Message based Interface Encoder/Decoder Module Stack Error Handling Message based Interface data structures and initialising these data structures with the appropriate values. The Finite State Machine (FSM) is the most important part of the Stack and is responsible for maintaining the state of all the connections managed by the stack. The FSM module can be organised as a collection of state-event functions. Then, based on the state of an connection and the type of signalling message received, the appropriate action can be invoked. signalling protocol entities communicate with peer entities by means of messages called signalling messages. On reception of such an message, the message is first decoded by the Decoder module and validated before appropriate action is taken. If an error occurred in the signalling message, the error handler is called. The Encoder module, on the other hand, takes appropriate parameters as input and forms these well defined signalling messages to be sent to a peer signalling entity. The Error Handler module contains procedures to be followed when an error is detected. Two types of error processing need to be offered. The first would be an error handling routine to handle protocol related errors and the second would be a handler to take specific action based on signalling related errors. Lastly, the Control module will be central to the operation of the Stack. This module will be responsible for ensuring that all the different modules can operate independently of one another by providing a suitable common interface between these different modules. Apart from these components, management related modules could be developed and provided [13]. 4. TEST BED TO BE IMPLEMENTED This section will focus on two implementations to illustrate how switching could be accomplished in a network by means of signalling. 4.1 LOGICAL IMPLEMENTATION The sequence of events that would occur in a typical network in order to demonstrate switching will be illustrated in Figure 5. V Fig.4 High-Level Architecture of Stack Besides the interfaces in both the Served layer and the STC layer that needs to be developed, the main Stack will consist out of the following components: A B C ATM VC / ATM Network ATM VC ATM VC W X The Initialisation module will typically only be used once at system start up to initialise the signalling stack. This can be accomplished either through Application Program Interfaces (APIs), or through embedded function calls. It would typically entail allocating memory for various Fig.5 Logical Test Bed Implementation Z Y

5 The diagram shows a number of users, all connected to various concentrators. All of these users are considered to be users. The concentrators are also connected to each other via the ATM network, which consists of ATM and switches. Each concentrator and its accompanied users are located at different physical locations. An signalling protocol entity is located on each concentrator and switch. Suppose users A, B and C needs to transfer traffic to users W, X and Z respectively. Since users A, B and C are connected to the same concentrator only one ATM VC is required between its concentrator and the ATM network. But three individual ATM VCs are required between the ATM network and the concentrators connected to users W, X and Z. Each initiating user would then request its bandwidth requirements to each concentrator and switch in the path, via signalling. If the ATM VCs can support the bandwidth requirements, signalling would then establish connections inside the ATM VCs to connect the corresponding users together. contained in the same destination end-point, they are logically considered to be located at different physical locations. The second end-point will thus provide the illusion of being a number of separated end-points with its own users. The provisioning of these channels will be done with signalling. An signalling protocol entity will be provisioned in each end-point as well as in the WUGS. 4.3 ROUTING One issue that has not been addressed in this paper is the routing concern. It was stated earlier that signalling does not contain a routing protocol, thus any routing mechanism implemented will be implementation specific. It was also stated that two independent network operators could be responsible for the individual ATM and networks. Two questions that needs to be addressed are, a) How could an network operator obtain a complete topology description of its network and b) how could an routing protocol be implemented in an network. The figure below shows a possible solution to the above-mentioned questions. Thus, one incoming ATM VC, from users A, B and C, would contain three multiplexed channels. The traffic contained in that ATM VC will then be separated by means of switching, in the ATM network, to direct the correct traffic to its corresponding destination. 4.2 PRACTICAL IMPLEMENTATION As discussed earlier in this paper signalling messages will be transported over, in the same ATM VC as the traffic. This also simplifies the implementation by not having to be concerned about the complexities involved in utilising two AALs, and AAL5. signalling messages were assigned a dedicated channel in each ATM VC. In our implementation we used CID 8 for this purpose. Figure 6 shows the practical test bed, which will be implemented. End-Point Entity Multiplexer Single Path SPC Switch WUGS SPC Fig.6 Practical Test Bed Implementation Multiple Path End-Point Entity Multiplexer The first end-point, on the left of the diagram, will be implemented on a PC and will consist of a few user instances, an signalling entity and an concentrator module. The concentrator module would multiplex the user traffic unto a single ATM VC, also referred to as an path. Once the traffic is transported to the switch, the switch will then perform the switching functionality and forward the traffic to the second end-point using multiple paths. Multiple paths are necessary, because although all the destination user instances are Fig.7 Implementing a Call Agent in the Network The network topology consists of and ATM switches. All the switches are connected to a centralised Call Agent. The Call Agent could poll the various switches on a regular interval to obtain certain information from the network. This could typically be information relating to the congestion of specific ATM VCs, which will enable the Call Agent to manipulate the routing tables of selected switches. By doing this, the Call Agent could prevent the over subscription of certain ATM VCs and thus prevent congestion from occurring in the network. The switches, in turn, will need to have a management platform available which the Call Agent will be able to interface to. The Call Agent would also gather statistics from the network that could be used for billing purposes and fault management. 5. CONCLUSIONS

6 switching was explained to be more efficient that simple trunking, but to enable to be a viable technology in real networks, it needs to be accompanied with a signalling protocol. signalling was standardised to address this need. The scope and limitations of signalling was examined in this paper. Some of the most important aspects of signalling, such as, which transport mechanism to use and how priority queuing could be used to minimised the delays incurred during signalling was discussed. A high-level architecture for the signalling stack was also proposed. Finally, a test bed implementation was looked at both from a logical and a practical implementation point of view. This was done to demonstrate how switching would be accomplished. The role and importance of signalling was also discussed in the test bed implementations. Finally, a Call Agent was proposed to address to routing problem in signalling. Although the Call Agent implementation is not within the scope of the current research, it could form part of future research. Access Networks, Proceedings of the Fifth IEEE Symposium on Computers & Communications 2000 [13] Nishit Narang, Sumit Kasera, ATM SVC Signaling Stack: Implementation Guidelines, URL: 6. REFERENCES [1] A CopperCom Technology White Paper, Mastering Voice Over DSL: Network Architecture, URL: [2] ATM Forum White Paper, Speaking Clearly with ATM - A practical guide to carrying voice over ATM, URL: tml [3] NTT DoCoMo, At the Core of 3G Mobile, IEEE Communications Magazine, Vol. 38 No. 12 December 2000 [4] Alcatel Technical Paper, Voice over DSL Quality Study, URL: DSLqos.jhtml [5] ITU-T, I B-ISDN ATM Adaptation layer specification: Type 2 AAL, September 1997 [6] Sven Shepstone, Neco Ventura, Andre van Zyl, A Gigabit Switch for Low Variable Bit Rate Services, September 2001 SATNAC Conference [7] Sven Shepstone, Neco Ventura, Active Switching Node to Support 3 rd and 4 th Generation Voice Services in Fixed ATM Networks, September 2002 SATNAC Conference [8] Goran Eneroth, Gabor Fodor, Gosta Leijonhufvud, Andras Racz, Istvan Szabo, Applying ATM/ as a Switching Technology in Third-Generation Mobile Access Networks, IEEE Communications Magazine June 1999 pg [9] Sumit Kasera, Pitambar Sahoo, Anil Sinha, Kuldeep Singh, : A White Paper, URL: [10] ITU-T, Q AAL Type 2 Protocol Capability Set 1, December 1999 [11] Shi Lu, Richard Thompson, Improving VTOA Transmitter Performance, Globecom 2001, San Antonio, Texas [12] Istvan Szabo, Sandor Szekely, Istvan Moldovan, Performance Optimisation of for Supporting Soft Handoffs in UMTS Terrestrial Radio