INTERNET-DRAFT G. Fodor Document: draft-fodor-intserv-wireless-issues-01.txt F. Persson Expires: July 2002 B. Williams Ericsson January 2002 Application of Integrated Services on Wireless Accesses Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This document is an individual submission to the IETF. Comments should be directed to the authors. ABSTRACT We consider a scenario wherein a QoS enabled application uses an Integrated Services based API (IntServ). Where the network interface is over a cellular wireless access network, we examine the feasibility of the radio management function to establish appropriate spectrum efficient bearers to support the requested service. We conclude that only the Controlled Load Service can be reasonably supported over a cellular wireless access network with the required spectrum efficiency. However, even for the CL service, some additional service parameters are needed in the IntServ model to allow the derivation of the necessary wireless parameters for spectrum efficient operation. Fodor, Persson, Williams [Page 1] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses 1 Table of Contents 1 Table of Contents 2 2 General Background 2 3 Wireless Network architecture and characteristics 3 4 Examination of Existing Integrated Services 6 4.1 Guaranteed Quality of Service (GQoS) 6 4.2 Controlled Load 7 4.3 Null Service 8 5 Conclusions 8 6 Security Considerations 9 7 IANA Considerations 9 8 Appendices 9 Appendix A Transmission Model for Wireless Network 9 Appendix A.1 Overview 9 Appendix A.2 Radio Channel Characteristics and Performance Optimization 12 Appendix A.3 Radio Channel Parameters 13 Appendix A.3.1 Traffic Descriptors 14 Appendix A.3.2 QoS Parameters 14 9 References 15 10 Author's Addresses 16 11 Full Copyright 16 2 General Background One of the key objectives in the evolution and standardization of various radio access technologies (as in 3G [3G-1, 3G-2]) is to offer QoS support for a variety of services, including IP services, while considering the critical aspects for wireless networks of optimizing spectrum efficiency. In an end-to-end path containing one or more wireless links, it is expected that the wireless links will be the most critical for QoS delivery. Thus, the end-to-end service will be mainly influenced by the suitability of the provided wireless link characteristics. One of the challenges in the design of QoS services, such as those specified by IntServ, is that they must be able to be supported effectively by a variety of link layers and QoS mechanisms, including those used by wireless access networks. It is believed that platforms supporting multiple interface types will provide generic (non-interface specific) QoS APIs to applications. Since an IntServ based API is expected to be widely available to application developers on some of these platforms, it is useful to ensure that spectrum efficient radio services can be provided for applications requesting QoS through an IntServ API. Fodor, Persson, Williams [Page 2] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses In this draft we focus on an IntServ enabled application/host requesting QoS through such an API, and investigate the feasibility of managing radio resources to provide spectrum efficient support for the requested service over a wireless access network. The draft is organized as follows. Section 2 describes the all-IP 3G architecture and the main functionalities of the wireless network elements. It also summarizes the aspects of radio access networks (described in detail in Appendix A) that are important to consider when evaluating the suitability of radio bearers to support the requested service. Section 3 examines the feasibility to provide appropriate spectrum efficient radio bearers for the individual IntServ Services and identifies some key issues for operating over a wireless access network. The Appendix develops a model for the radio access network, and describes the radio mechanisms available to vary the QoS characteristics of the radio bearer services. It also details a set of radio parameters that may be available to control those mechanisms. 3 Wireless Network architecture and characteristics In this context, the term _all-IP architecture_ is often used to indicate that the IP layer is present in the mobile terminal and that the user may establish an IP connection to other IP endpoints through an IP network. In an all-IP network, a typical scenario includes the following elements: o The terminal equipment (UE) is considered to include the physical device connecting to the Wireless Access Network (WAN), as well as any additional equipment provided with service through this connection (e.g. a mobile terminal and a PC attached to it with a bluetooth connection). o The Wireless Access Network (WAN) that consists of base stations (BS), base station controllers BSC, (also referred to as radio network controllers, RNC) and possibly other nodes responsible for mobility management, location management, etc. The WAN connects to the external IP Network (e.g. The Internet) through (a) gateway node(s) (WAN GW). It is important to note that the WAN appears to the UE as a L2 network; and is designed and optimized for the transmission of radio packets. + + +--++ Terminal /-----\ | | Equipment / \ | UE| | IP | | | | Network | +---+\ --------------------- | |- +----+ \ / \ | | \--| UE | \/ WAN \ /\ / +----+ /\ /--\ /--\ GW \/ \---+-/ Remote Fodor, Persson, Williams [Page 3] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses / \ | | | +--+ /\ | Terminal / \| B | B | | |/ \ sss / | | | /| | \ sss | /--\-/--\-/--\ / +--+ | sss | | | ccc| |ccc | | | B | ccc| B |ccc | | | | ccc| |ccc | | \--/-\--/-\--/ | \ | | | / \ | B | B | Radio / \ | | | Access / Wireless \ \--/ \--/ Network / Access \ / Network \ / --------------------- ccc sss B = base ccc = base station sss = application Station ccc controller sss server Figure 1 A simple diagram showing UE, WAN-GW, and IP Network chain The radio packet delivery service (associated with a specific set of traffic and QoS characteristics) that is provided by the WAN is often referred to as the radio access bearer service (RABS). For instance, a streaming RABS may mean a radio packet delivery service provided by the WAN that provides bounded delay and limited packet loss ratio. The set of radio level entities that realize the RABS are referred to as radio channels or radio bearers. A simplified diagram of the communication within an all-IP wireless network is shown by Figure 2. An application in the UE uses an IntServ based API to request QOS service. With the 3G all-IP network, the terminal equipment has the responsibility for identifying the radio bearers that it needs, and how it will use them. Thus, it is responsible for initiating the radio bearers between the UE and the WAN GW. Here we are examining a scenario where the UE makes use of the IntServ service information in determining the appropriate radio bearers to establish. +----+ +----+ |Appl|<----+ |Appl|<----+ | | V | | V +----+ +-------+ +----+ +-------+ |TCP/| |IntServ| |TCP/| |IntServ| |UDP | | Module| |UDP | | Module| +----+-+-------+ +-----+ ,-. +----+-+-------+ | IP |<---------------->| IP |<-------->| IP | +--------------+ +--+--+ +--+--+ ,---. +--------------+ | | ,---. | |IP|<>|IP| | | | | | | Radio L2 |<------->|R2+--+ +--+L2|<-------->+ L2 + | | | | | |T2|<>|T2| | | | | | Fodor, Persson, Williams [Page 4] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses +--------------+ | | +--+--+ +--+--+ | | +--------------+ | Radio L1 | \ / |R1|T1|<>|T1|L1| \ / | L1 | +--------------+ `-' +--+--+ +--+--+ `-' +--------------+ UE Radio RNC/ WAN IP Remote (terminal Network BSC GW Net User equipment) \ / -----v----- RAN Figure 2 Simplified End-to-end 3G All-IP Model The WAN provides radio access bearer services to support the layer 2 connection between the UE and the WAN GW. The characteristics of these bearer services are dependent on the wireless mechanisms, and can be markedly different from bearer services in traditional wired networks. The appendix describes a model for the radio access bearer services that can be provided by a wireless access. It then goes on to show how parameters of the radio bearer service can be used to change its characteristics. Using the model as defined in the appendix, it is clear that different radio bearer services can be provided, resulting in quite different characteristics of QoS, service costs and service behaviors. A consequence of this flexibility is that sufficient detail about the applications' traffic and service requirements must be known in order to determine the appropriate parameter settings to enable service optimization.. In wireless networks, the physical resources (e.g. frequency spectrum, transmission power or time slots) are typically scarce. The transmission is therefore optimized in order to utilize the resources as efficiently as possible. For example, one optimization is to maximize the total throughput [kbps/MHz/cell] of a system for a given service quality. Another optimization is to maximize the number of admitted users for a given service quality level [Erlang/cell]. The number of admitted users is essential for the revenues of the operators. On the other hand, when increasing the number of users, the overhead and interference levels are increased. Since the resources are shared between the users, there is a trade-off between the number of users in the cell and the bit rate and service quality level that can be offered to each user. The realization offering the best resource utilization and service cost is normally used. Since service cost and spectrum efficiency are crucial in wireless networks, the radio bearer services selected should be aimed to just meet (not to exceed) the actual QoS requirements. Fodor, Persson, Williams [Page 5] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses The appendix identifies a detailed range of characteristics that could be controlled in radio bearer services. For evaluating the suitability of radio bearers to support current Integrated Service definitions, we summarize the characteristics of the radio bearer services as follows: o A radio bearer service (and the corresponding radio channel) providing minimum delay would not have retransmission at the radio layer, and could have minimal error correction codes and interleaving. This could result in high bit error rates. o Reliability can be improved through several parameters. Radio channels typically have a trade-off in one or more characteristics, such as delay (e.g. from interleaving, FEC, ARQ), spectrum usage (e.g. from FEC) or number of users that can be supported (e.g. from signal power). Optimal use of the network resources needs sufficient information about the service usage to identify the most appropriate setting of the range of parameters. For example, a telephony application using an AMR codec would be well supported by a radio bearer that provides minimal delay, even if the error rate is quite high. Alternatively, another application may require low delay, but may not tolerate a high error rate, so the power level and FEC may be different. 4 Examination of Existing Integrated Services Integrated Services enhance the single best-effort service paradigm by introducing multiple services. IntServ allows the network nodes to perform explicit resource management at the IP level by allowing applications to characterize their resource requirements. IntServ requests allow the nodes to exercise admission control and traffic control and thereby provide end-to-end QoS provisioning and resource allocation. An end-to-end service may traverse a number of QoS domains. We are only concerned here with the feasibility to establish spectrum efficient (L2) radio bearers to support the service request as it applies to the radio bearer link (i.e. between the UE and the WAN GW). Issues with support of the service requirements over the end- to-end path are outside the scope of this document. 4.1 Guaranteed Quality of Service (GQoS) The Guaranteed Quality of Service [RFC2212] Integrated Service is defined to provide a service of guaranteed bandwidth with a bounded queuing delay. The end-to-end behavior provided by a series of network elements provides an assured level of bandwidth that, when used by a policed flow, produces a delay-bounded service with no queuing loss for all conforming packets (assuming no failure of network components or changes in routing during the life of the flow). Fodor, Persson, Williams [Page 6] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses The GQoS service expects constant bandwidth and a low loss rate, since loss is normally expected to result from congestion which GQoS would minimize since it reserves resources. For wireless networks, the bit error rate is much higher, and is influenced by many different factors including path loss, fading and interference (see Appendix A). These are most likely to change when the user is moving, but may also vary even when the user is stationary. It is difficult to meet both the low loss and constant data rate expectations of the GQoS service given the variation in radio channel performance. GQoS permits the user to increase the bandwidth of the service to control the queuing delay, as this was typically the only variable in end-to-end delay of a best-effort network. For public cellular wireless networks, it is important to be able to provide service to as many customers as possible, so service charges are typically dependent on the service. Thus, although it may be technically feasible for the UE to request greater bandwidth than the traffic needs to meet the service requirements, it is not cost effective to do so, and thus must be avoided. A wireless network is one which can have a wide variation in delay dependent on the radio bearer service (consider the transmission model described in the appendix). Information about the delay and delay variation for the service can be provided to the end host. The worst case delay variation (the difference between the minimum delay the network can support, and the maximum that could occur) for radio bearer services is dependent on the bearer characteristics (e.g. retransmissions), and thus can be extremely large. Hence, the reported delay variation is not beneficial in determining the required bandwidth. Finally, there is insufficient information in the GQoS service to determine appropriate settings for the radio parameters (e.g. can the application tolerate high bit error rates). 4.2 Controlled Load The Controlled Load (CL) Integrated Service is defined to provide a service approximating that of a lightly loaded network. The service should provide that a very high percentage of transmitted packets will be successfully delivered by the network, and the transit delay of most packets will not greatly exceed the minimum transit delay of the routing vector. Applications with different service requirements must be provided bearers with suitable characteristics for that service. With the wireless network admission controls, a radio bearer can receive characteristics similar to that of a lightly loaded wireless network for that type of radio bearer. Fodor, Persson, Williams [Page 7] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses However, the CL service does not provide sufficient detail to identify the most suitable bearer, and thus the expected characteristics of _lightly loaded_ which are aimed for are not known. 4.3 Null Service The Null Service (NS) allows applications to identify themselves to network QoS policy agents rather than requiring them to specify resource requirements. It is important to note that the policy must be identified per user for each application. Since the selection of radio bearers is performed from the UE, the policy agent could be provided there. However, this places the burden for managing/configuring this agent on the end user. Furthermore, although many parameters for the radio bearer services could be identified based on the application, there are additional parameters that need more information about the actual specifics of the application session (for example, within one application, the choice of codec may determine the required bit error rate). 5 Conclusions In this report we considered the problem of allocating network resources in a scenario where IntServ enabled applications/hosts access an IP network through a wireless network. This is a meaningful scenario that supports the separation of the applications from the specific interface technology, in this case, a wireless access. In such a scenario the scarce wireless resources need to be allocated in the most effective manner to deliver service to the applications. While the specific radio related traffic descriptors and QoS parameters may vary between wireless technologies, we argue that additional information is necessary to achieve spectrum efficient service. The considerations in providing appropriate spectrum efficient radio bearers for the existing Integrated Services can be summarized as follows: Guaranteed Quality of Service o It is difficult to meet the expectation of low delay and low loss in the wireless environment given the variation in radio channel performance. o Increasing the bandwidth to reduce the queuing delay would be relatively ineffective, but would disproportionately increase the user's service costs. o Reported delay variation for wireless networks is of no benefit in determining the required bandwidth. Fodor, Persson, Williams [Page 8] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses o Sufficient information to allow appropriate setting of the wireless parameters (e.g. acceptable packet loss ratio) is not available. Controlled Load o Additional information is required to determine the appropriate characteristics expected from the network. Null Service o Requires the implementation of the QoS policy agent within the UE. o Additional information besides the application/sub-application would be required for the ideal setting of the wireless parameters. Therefore, we conclude that only the Controlled Load Service can be reasonably supported over a cellular wireless access network with the required spectrum efficiency. However, even for the CL service, some additional service parameters are needed in the IntServ model to allow the derivation of the necessary wireless parameters for spectrum efficient operation. 6 Security Considerations None 7 IANA Considerations None 8 Appendices Appendix A Transmission Model for Wireless Network Appendix A.1 Overview When considering the effect of a wireless access network on an IP service, it is useful to have a model that describes the characteristics of the transmission over the wireless access network. This model is described below. Following the development of the model, various parameters are identified that may be available in wireless networks to control the characteristics of individual radio bearer services that can be offered to the higher layer. Conversely, this means that IP service requests need to be sufficiently granular to enable appropriate setting of these controls. From a modeling point of view, the WAN is an L2 network. A queuing model for a WAN (L2) network (Figure 3) is somewhat different from Fodor, Persson, Williams [Page 9] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses the _traditional_ queuing models (e.g. as used to model an output port of a router). The formulation of a queuing model for the L2 is important, because it serves as the basis for the discussion on the effects of different mechanisms in Section A.2. To develop the model, it is useful to follow the steps of transmitting an IP packet over a wireless access network. We will use the uplink direction (i.e. from the TE to the CN and the server), but the major considerations are valid in the reverse direction as well. At the radio layer we use the term _radio service data unit (SDU)_ to refer to the piece of data that is interchanged between the IP and radio layers. That is, an IP packet that is handed over to the radio layer (L2) for transmission in the MT (in uplink) becomes a radio SDU. The radio SDU is thereafter segmented into smaller radio packets. The use of smaller radio packets helps to reduce the probability of bit errors in a single radio packet. The optimum radio packet size is a trade-off between radio level overhead (radio packet header, CRC, number of retransmissions, etc.) and the packet error probability. The radio packets are transmitted over the air interface (AI) to the base stations (BS) and then through the radio network controller or base station controller (RNC/BSC) to the WAN GW. We note that a UE may be in contact with one or more BS's at a time e.g. in CDMA networks. The issues related to handover and soft hand-over can be found elsewhere (e.g. [SHO]) and are out of the scope of this draft]. Input IP Packet segmented into radio frms +-----------------------+ |+----++----+ +----+ | Output IP || Rad|| Rad|..| Rad| | Packet || Frm|| Frm|..| Frm| | /\ |+----++----+ +----+ | || +-----------------------+ || || || prep of radio || Rate || frames || Control || eg. FEC, CRC || | || & radio header || V || \/ /----\ || +-----------------+ | R | +---+ +--------------+ |AAA,BB,CCC,DDD,..|--->|(Chnl |--->| |--->|#3 Reassembly | +---------------v-+ | Rate)| +---+ +------v-------+ ^ v \--v-/ Fixed | | v v Trans | | #1 v Delay | | #2 | Fodor, Persson, Williams [Page 10] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses +-----------------------------------------+ Radio Packets scheduled for re-transmission (ARQ) #1 _ loss due to radio buffer overflow #2 _ loss due to transmission effects (eg interference,path loss, noise) #3 _ packets received (possibly with bit errors) Figure 3 Simplified model of the transmission of a radio packet The radio packets are queued for transmission over the air interface. The L1 transmission over the air is modeled as a queue server of capacity R bit/s that corresponds to the actual physical bit rate of the radio channel. The radio channel is lossy, meaning that radio packets may get lost (due to interference, path loss, etc), or be partially damaged by single or consecutive bit errors. Depending on the signal to noise ratio, the bit error ratio over the wireless link could be up to the order of 10E-02 or even higher. Additionally, radio packets face a significant transmission delay over the radio interface independent of the channel rate. This is taken into account by introducing a fixed transmission delay in the model. Radio packet retransmission (ARQ) is available which is indicated by the reinsertion of radio packets in the queue, but we note that for real time services this mechanism typically is not used. In Figure 3 it is assumed that all radio packets belonging to a single IP packet share the same queue (i.e. the same radio channel). Furthermore, the model also assumes that the radio channel capacity R is constant, which is not necessarily an accurate assumption, since the radio channel capacity (in terms of transmitted bits per second) may decrease/increase due to rate control mechanisms at the radio layer. Rate control mechanisms play an important role in the WAN as they allow the WAN to adaptively increase/decrease the radio resources associated with a radio channel for a specific user. Of course, from the IP layer perspective, the available rate varies due to rate changes at the radio packet level, as well as variation due to segmentation and retransmission. While this model is sufficient for examining the usability of the IntServ, it should be noted that in some cases it makes sense to assign different radio QoS classes for the different parts of an IP packet. For instance, this is the case for the popular voice codec Adaptive Multi-rate (AMR) codec [AMR], where the different segments (bits) of the coded speech samples represent unequally important pieces of information for the human ear. Therefore the different parts of a speech sample may be transported over different radio channels offering different error protection. This mechanism is often referred to as unequal error protection (UEP) [UEP]. Fodor, Persson, Williams [Page 11] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses In the WAN the radio layer is terminated. Parameters may control whether lost/erroneous radio packets are retransmitted or not. For real time traffic, retransmission may not be requested, and therefore the re-assembled IP packets may contain bit errors both in the payload and in the header. Appendix A.2 Radio Channel Characteristics and Performance Optimization In a wireless communication system the signals are disturbed by several reasons while propagating from the sender to the receiver. The most common reasons include path loss, fading and interference. Path Loss refers to the loss in signal power due to distance and indirect propagation paths (environment dependent). Fading is experienced when the signal propagates over multiple paths (fast fading) or is shadowed by e.g. hills and other obstacles. Since the experienced path and environment are changing not only during the movement of a user, but also when the user is stationary, the experienced path loss and fading typically are changing both in time and space. Interference results from noise and transmission signals of other users. There is a range of radio transmission mechanisms to combat these problems. The desired performance (delay, bit rate etc.) and cost in resources and complexity will decide which ones to use. Different applications and services have different requirements, and will require different mechanisms. While the environment cannot be controlled, the required signal-to-interference ratio, SIR, normally can be achieved _ either directly by increasing the signal power or indirectly by lowering the SIR requirement. The direct method of increasing the signal power is quite straightforward. When the SIR is increased the receiver becomes less sensitive to path loss, fading and interference. The drawback is that increased signal power for one user is experienced as increased interference to the other users nearby. An increased interference level leads to degraded quality (more bit errors) or, if the quality for each user should be maintained, lower system capacity (since the number of users then must be decreased). A well-balanced SIR must therefore be combined with some other mechanisms ensuring the reliability (bit- and packet error ratios). By adding redundancy to the radio packets, the SIR requirements can be decreased. One way of doing it is to retransmit erroneous packets until the required reliability is reached (ARQ _ Automatic repeat request), as shown in Figure 3. Errors are detected by a CRC (Cyclic Redundancy Check) code. Another way is to add redundancy in error correcting codes (FEC _ Forward Error Correction). ARQ is often used in combination with FEC. The advantage of ARQ is the low average signal power that is needed. The drawback is that the contribution from the fixed transmission delay (and partly also the queuing delay) will increase proportionally with the number of retransmissions. The advantage Fodor, Persson, Williams [Page 12] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses with FEC is the low transmission delay (a packet is transmitted only once). However, higher signal power is required (since fewer errors will be accepted over the channel), which leads to increased interference in the system. Further, the same high reliability as obtained by ARQ is not possible by only using FEC. The way the reliability is implemented (ARQ and/or FEC) affects both the performance (in reliability and delay) and number of served users. The correction capability of the error correcting codes can be improved by adding interleaving to the bit stream. Consecutive bits are then sent in a non-consecutive way, so that error bursts residing from fading on the channel are spread over the message. Isolated errors are easier to recover than errors residing in bursts. It will therefore be possible to better recover the radio packets with the unaffected bits. The degree to which the stream is interleaved is defined by the interleaving depth (i.e. the time interval over which the errors are spread). Larger interleaving depth will strengthen the effect, but leads also to longer transmission delay over the radio (a big part of the fixed transmission delay in Figure 3 can be derived from interleaving). Appendix A.3 Radio Channel Parameters Resource allocation for and QoS handling of the radio SDUs are typically done on a per-radio channel basis. The specific radio technology may support various QoS classes and QoS parameters to support QoS provisioning and efficient radio resource utilization. Typical radio admission control algorithms base the admission decision on the description of the traffic and the desired QoS (Figure 4). Radio Radio Network Network Interface Interface | | | | | /----\ | +------------+ | | Radio| | +-------------+ |Radio SDU in|------->| Netwk|------->|Radio SDU out| +------------- | ^ | (L2) | | +-------------+ | | \----/ | ^ Radio traff params-+ | | | are defined here | | +-- May have experienced: - Peak Rate | | - Bit errors - Avg Rate | | - Radio SDU losses - Max SDU size | | - Sequence errors Radio | Admiss +--Radio QoS defined here Cntrl - Radio SDU transfer delay [ms] - Radio SDU error ratio - Bit error ratio Fodor, Persson, Williams [Page 13] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses - Sequence integrity (deliv order) [y/n] - Delivery of erroneous SDUs [y/n] Figure 4 The scope of the radio network traffic and QoS parameters In the following we discuss some of the most common traffic and QoS parameters that are used at the radio layer for managing resources, exercise admission control, etc. Appendix A.3.1 Traffic Descriptors For real time traffic, the radio channel is typically characterized by a peak bit rate, a long-term average bit rate, and a maximum packet size. The traffic offered to the radio network needs to be conformant to all of these characteristics in order for it to receive the negotiated radio level QoS. If the packet sizes are fixed or within a strictly limited set of sizes, the radio network can be further optimized if the exact sizes and internal payload format are given. Appendix A.3.2 QoS Parameters The QoS parameters describe the expected performance of the network as measured at the egress point (Figure 4). Since packets can be lost, delivered out of order or with bit errors, it is reasonable to consider the following QoS parameters: o Radio SDU transfer delay [ms] Since the radio SDU delay in general is a random variable, the delay is usually characterized as the maximum delay for a high value (e.g. 95%) percentile of the distribution of delay for all delivered SDUs during the lifetime of a radio channel. The purpose of specifying this parameter is to quantify the delay that is tolerated by the application using the radio channel. This parameter may be used by the WAN e.g. to set the maximum number of allowed retransmissions (if any) and the interleaving depth. o Radio SDU error ratio This parameter indicates the maximum fraction of radio SDUs that can be lost or detected as erroneous. The SDU error ratio is usually defined only for conformant traffic, and is in radio networks an important quality measure (especially for services like speech). o Bit error ratio (BER) The bit error ratio (BER) indicates the fraction of the total number of delivered erroneous bits. It cannot be measured explicitly in the network, but is instead used as an implicit requirement, given that the requirement on Radio SDU error ratio Fodor, Persson, Williams [Page 14] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses is fulfilled. The BER requirement is used when deciding the CRC length (variable in radio networks). It can also be used for weighting of the coding rates between different radio bearers. This weighting cannot be derived from the Radio SDU error ratio if the radio packet sizes are not known beforehand, which is a common case. o Delivery of erroneous radio packets required [y/n] This parameter indicates whether radio packets with detected bit errors should be delivered or not. In the case of unequal error protection, this attribute should be set for each sub-flow. This parameter is needed for applications that can handle bit errors in the payload (e.g. voice codecs as AMR) in order to appropriately control the behavior of the WAN. o Sequence integrity (delivery order) required [y/n] This parameter indicates whether the radio channel must provide in-sequence delivery of the radio SDUs or not. o Traffic handling (packet scheduling) priority For some types of radio channels it can be advantageous to specify relative, rather than absolute, QoS guarantees for a radio channel. For instance, a _better than best effort_ type of radio channel may require a higher priority than a _best effort_ radio channel without specifying an absolute delay bound. In such cases the notion of the packet handling priority may be useful. When specified, the WAN uses this parameter for radio packet scheduling. 9 References [3G-1] Dahlman, E, Gudmundson, B, Nilsson, M, Skold, J, _UMTS/IMT-2000 Based on Wideband CDMA_, IEEE Communications Magazine, Vo. 36, No. 9, September 1998. [3G-2] Nikula, E, Toskkala, A, Dahlman, E, Girard, L, Klein, A, _FRAMES Multiple Access for UMTS and IMT-2000_, IEEE Personal Communications Magazine, Vol. 5, No. 2, April, 1998. [RFC2211] Wroclawski, J., "Specification of the Controlled-Load Network Element Service", RFC 2211, September 1997. [RFC2212] Shenker, S., Partridge, C., Guerin, R., "Specification of Guaranteed Quality of Service", RFC 2212, September 1997. [RFC2997] Bernet, Y., Smith, A., Davie, B., "Specification of the Null Service Type", RFC 2997, November 2000. [intdiff] Bernet, Y., Yavatkar, R., Ford, P., Baker, F., Zhang, L., Nichols, K., Speer, M., Braden, B. and B. Davie, "A Framework for Integrated Services Operation over Diffserv Networks", RFC 2998, November 2000. Fodor, Persson, Williams [Page 15] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses [diffarch] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z. and W. Weiss, "An Architecture for Differentiated Services", RFC 2475, December 1998. [AMR] Sjoberg, J, Westerlund, M, Lakaniemi, A, Koskelainen, P, Wimmer, B, Fingscheidt, T, Xie, Q, Gupta, S, _RTP Payload Format for AMR_ draft-ietf-avt-rtp-amr-10.txt, work in progress [UEP] Xie, Q, Gupta, S, _Error Tolerant RTP Payload Format for AMR_, draft-xie-avt-et-rtp-amr-02.txt, work in progress [SHO] Kempf, J., McCann, P., Roberts, P., _IP Mobility and the CDMA Radio Access Network: Applicability Statement for Soft Handoff:, draft-kempf-cdma-appl-01.txt, work in progress 10 Author's Addresses Gabor Fodor Ericsson Research Ericsson Radio Systems AB Torshamnsgatan 23 SE-164 80 Stockholm, Sweden Phone: +46 8 404 3084 Email: Gabor.Fodor@era-t.ericsson.se Fredrik Persson Ericsson Research Ericsson Radio Systems AB Torshamnsgatan 23 SE-164 80 Stockholm, Sweden Phone: +46 8 585 30818 Email: Fredrik.F.Persson@era.ericsson.se Brian Williams Ericsson AsiaPacificLabs Australia 33/360 Elizabeth St, Melbourne, Australia, 3000 Phone: +61 3 9301 4675 Email: brian.williams@ericsson.com.au 11 Full Copyright "Copyright (C) The Internet Society (date). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published Fodor, Persson, Williams [Page 16] INTERNET DRAFT Application of IntServ Services Expires July 2002 on wireless accesses and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Fodor, Persson, Williams [Page 17]