 |
|
|
|
Real-time, distributed multimedia applications demand sophisticated communications services, specifically quality-of-service multipoint-to-multipoint delivery, from the underlying network. The subject of considerable research, developing effective means for providing these services has proved challenging, even in the context of stationary wireline networks. We are currently investigating the problem of supporting multimedia applications in networks in which both the endpoints and the switches forming the network infrastructure are mobile and communicate over wireless links (e.g., multihop, mobile packet-radio networks). Node movements and wireless communications result in highly variable network connectivity and transmission capacity. Signal degradation or loss over a link may result when two nodes move apart, when higher-power transmissions from other nodes interfere, or when obstacles or external noise sources are present in the environment. Operating under these conditions, the network algorithms must be able to react quickly, efficiently, and correctly, even with imperfect knowledge of network state, if they are to succeed in providing the high throughput and low delay, jitter, loss, and corruption needed for multimedia transport. Our approach to supporting multimedia applications in mobile wireless networks combines adaptation to and control of the network's variability and focusses on link-layer and network-layer algorithms. These algorithms are designed to function together as a modular, parameterized system that can be tuned appropriately for the expected network environment and the desired performance. The system comprises the following components:
Self-Organizing Hybrid Routing Network. Switches and endpoints autonomously form a virtual, hierarchical routing network, through clustering. This network organization can accommodate thousands of nodes and combines elements of both cellular and packet-radio networks. Endpoints affiliate themselves with switches according to the perceived quality of the links between them, forming cells, each with a single switch within one hop of its affiliated endpoints. Switches group cells into clusters, clusters into higher-level clusters, and so on, forming a virtual, multihop network within each level. Criteria for splitting and coalescing clusters include minimizing the volatility of inter-cluster connectivity and limiting the number of component clusters.
Adaptive Link Control. Each pair of nodes connected by wireless link executes an adaptive link control procedure which helps to stabilize the link's quality. Specifically, a node measures parameters of received transmissions (e.g., bit-error rate) and communicates these to its neighbor who then adjusts its radio parameters (e.g., transmitter power) to boost the quality of the link to the desired level. The available measured and adjusted parameters will depend on the capabilities of the specific radio. The adaptive link control algorithm has two modes of operation: real-time (milliseconds) requiring a specific slotting of the channel and short-term (seconds) providing control on a packet-by-packet basis.
Hierarchical Location Management. Each node has a roaming level which is specified which respect to the clustering hierarchy and which implicitly defines a roaming cluster. The node generates a location update only when it moves outside of its current roaming cluster. Paging is required to locate a moving node within its current roaming cluster. For the purposes of routing, outside of the roaming cluster, the address of a node is advertised as the address of its current roaming cluster. Only within the roaming cluster must the complete current address be known. Thus, information about changes in a node's lcoation propagate no further than the cluster in which the changes occur.
Multi-Level Quality-of-Service Routes. A cluster's service characterization for routing depends upon the services provided by its component clusters, which in turn depends upon the constituent nodes and links, and hence may be highly volatile. Therefore, a combination of abstraction of advertised services of component clusters together with actual measured services across the given cluster is employed to provide an accurate statistical service characterization. This service characterization also includes the degree of mobility of nodes within the cluster. Route selection accounts for application service requirements and cluster service constraints, and favors minimum-hop routes through less mobile clusters. Routes are specified relative to the clustering hierarchy and at multiple levels, so that a lower-level portion can be modified in response to network changes while the higher-level portion remains intact.
Dynamic Multipoint Virtual Circuits. Each stream-oriented session attempts to acquire a dynamic virtual circuit with the necessary resources, on demand. The virtual circuit, like its underlying route, is multilevel with respect to the clustering hierarchy. Multipoint virtual circuits possess a dynamic branching facility that enables mobile endpoints to join multicast sessions rapidly. To minimize session service interruptions, virtual circuit management includes several types of repair mechanisms: local handoffs that do not require routing intervention; rerouting a portion of a virtual circuit; establishing a multipoint virtual circuit between a source and the predicted locations of a mobile destination, with one active branch at a given time; establishing multiple virtual circuits between a source and destination, when intervening nodes are highly mobile and accurate location prediction is not possible, with all paths active simultaneously.
In this presentation, we describe the algorithms composing each component of the system, paying particular attention to routing and dynamic virtual circuit management. We also present selected results from our ongoing simulation study of these algorithms and their interactions. Implementation of the system in an experimental packet-radio network is planned for the coming year.
# This project is funded under the DARPA (Defense Advanced Research Projects Agency) GloMo (Global Mobile Information Systems) program. The other members of the project team are Marcos Bergamo, Regina Rosales-Hain, and S. Ramanathan.}
New data standards are being negotiated and proposed to allow for coexistance of data services with voice on the same RF channel. New Point to Point protocols and the Radio link protocols are being proposed to support data services using IS-95. The proposed usuage of these data services are faxing, internet access, file transfers etc.
The data services related usage has different characteristics than voice users. Therefore their use of RF resources is different. Thus there can be situation where a a single data user may use up more than a single voice user's worth of bandwidth. Thus a data user may displace more than a single voice user.
In general the end user expects that the data services be less expensive than voice usuage and expectes to pay less or about the same as voice access. The service provider on the otherhand has to characterize the usage for each kind of data users and price services according to the usuage.
In this talk we present a framework to model the co- existence of the data and voice users on a single RF channel. We present results that describe the effective usuage of resources as a function of adding data users.
The advances in low bitrate video coding technology have led to the possibility of delivering video services to users through band-limited wireless networks. Both ITU-T/SG15 and ISO-MPEG4 are working to set standards for very low bitrate video coding. Recently, ITU-T/SG15 finished the first draft recommendation of H.263 which targets the transmission of video telephony through Public Switched Telephone Network (PSTN) at data rates less than 64 kbit/s. The expert's group is starting to adapt H.263 for wireless applications because the low bitrate makes it well suited for band-limited wireless networks.
Real-time video services require high reliability with a low bounded time delay and a reasonably high transmission rate. Radio channels on the other hand are error-prone, time-varying and band-limited. Proper error control is necessary to obtain acceptable quality video transmission. Traditionally, forward error correction (FEC) codes are used for real-time services because they maintain a low bounded delay. Wireless channels are time-varying. FEC codes can be chosen to guarantee certain quality of service (QOS) requirements for the worst channel conditions. However, this causes unnecessary overhead and reduces throughput during periods of good channel status. Hybrid automatic repeat request (ARQ) error control schemes make use of both forward error correction and retransmission in order to achieve high throughputs and high system reliability. Their adaptability to fluctuating channel conditions makes them attractive to use for the transmission of compressed video over band-limited wireless channels.
In this paper, we propose a new hybrid ARQ error control scheme based on the concatenation of a Reed-Solomon (RS) code and a Rate-Compatible Punctured Convolutional (RCPC) code. The transmission of H.263 coded video over an ATM based wireless network with error protection provided by the proposed hybrid ARQ is investigated. We limit the allowed maximum number of retransmission attempts to bound the time delay within an acceptable range for real-time video services. The performance of the proposed hybrid ARQ scheme and the conventional hybrid ARQ schemes are compared for H.263 video transmission.
The proposed scheme combines the advantages of hybrid ARQ and the power of RS-RCPC concatenated codes. The inner RCPC code, whose rate is based on the channel conditions, is able to provide each (re)transmitted packet with certain error correction capability (like type-I hybrid ARQ). It is generally reliable and the information can be recovered from the initial transmission when the channel is not in long error bursts. This reduces the frequency of retransmission, thereby maintaining a low time delay wit ed! by the accompanied RCPC code. This is very important for time varying wireless channels, where an error burst or a header error might wipe out most of the initial transmission yet leave the retransmission relatively error free. Furthermore, the RS parity blocks can be combined with the initially transmitted information blocks to form a powerful RS-RCPC concatenated code to correct errors when error correction has failed for every single packet (like type-II hybrid ARQ). Simulation results show that the proposed hybrid ARQ scheme can significantly improve the quality of video transmission because of its ability to effectively adapt to the varying channel status and because of its powerful error correction capability.
Different broadband services require different amounts of bandwidth and have different priorities. For example, a connection for visual communications will in general require more bandwidth than one for data communications, and a voice connection will in general be of higher priority than either a data or a video connection. In response to these varied demands, the network designer may choose to assign different amounts of bandwidth to different types of traffic. The motivation for such an approach stems from the desire to support different kinds of multimedia services with a reasonable level of performance and without letting the demand from any one type shut-out other types of services. The challenge for the designer is to come up with techniques that are able to balance the needs of the various applications with the need of the system to accommodate as many connections as possible. This task of providing guaranteed quality of service with high bandwidth utilization while servicing the largest possible number of connections can be achieved through a combination of intelligent admission control, bandwidth reservation and statistical multiplexing.
Supporting real-time VBR video along with voice and data over bandwidth-constraint networks continues to be formidable problem. The difficulty arises because VBR video is unpredictably bursty and because it requires performance guarantees from the network. While resource reservation schemes work best for CBR traffic, there is no consensus on which strategy should be used for VBR traffic. On one hand, since real-time VBR traffic is delay sensitive, a resource reservation scheme seems to be the right choice, on the other hand, because VBR video is bursty, if resources are reserved according to peak rates, the network may be under- utilized if the peak-to-average rate ratios are high. These two opposing characteristics have resulted in a common belief that it is unlikely that performance guarantees can be provided to such bursty sources with very high network utilization. This is the problem we address in this work, that is, can performance guarantees be provided to VBR video without significantly under-utilizing the bandwidth and can this be done in conjunction with minimizing the maximum blocking probability for voice and data connections?
Our solution to this problem consists of three parts: (1) use a multi-resolution joint source-channel video codec, (2) provide connection lifetime reservation for real-time video connections with optimum bandwidth utilization, and (3) partition the available bandwidth in a manner that ensures that the maximum blocking probability for voice and data traffic is minimized.
From a connection’s perspective, we advocate the use of a multi-resolution subband video codec (or a segmented region- based video codec) [1]. We suggest reserving the peak bandwidth for the primary subband (or region) while letting the secondary and tertiary subbands (or regions) compete dynamically for bandwidth (see Figure 1). A potential problem with this approach is that most of the time the actual amount used by the primary subband is far below the amount reserved for the peak. We concentrate on this aspect and develop techniques to completely use all of the reserved bandwidth.
From the system’s perspective, we develop a simple yet effective algorithm that partitions the available bandwidth in a manner that minimizes the maximum blocking probability for voice and data connections while providing guaranteed QoS to VBR video connections (see Figure 2). It should be noted, that even when the distribution of the different traffic types is given, finding the optimal partitioning of the bandwidth is a very difficult task, and for the general case can be modeled by an NP-complete graph coloring problem. The intractability of finding the optimum is present already in the simplest situation when the traffic consists of voice connections only and the statistics of the offered traffic are completely known. However the problem becomes even more difficult when the wireless network is carrying integrated non-homogeneous traffic, a situation occurring naturally in the case of wireless multimedia networks. In this case estimating the blocking probability of connections and its application in resource allocation strategies are further complicated for two fundamental reasons:
We propose a solution for the allocating transmission resources among different traffic types under incompletely known conditions. This when combined with our proposal for using multiresolution video and bandwidth reservation with intra-frame statistical multiplexing, our solution has the following main properties:
The goal of the InfoPad project is to demonstrate a complete system solution to accessing and manipulating multimedia data (text/graphics, video and audio) from backbone network based information and compute servers. The user device, the InfoPad, is being designed to be as light weight, low power and low cost as possible, essentially being set by the display requirements. In order for the user to appear to have large amounts of local storage and a high performance computing platform, it is necessary to optimize the wired and wireless communications links between the pad and the servers and to provide the networking software which can support the pads mobility as well as to continuously adapt to the varying quality requirements of the multimedia data. This all to be accomplished under the constraint of minimum power consumption in the portable units.
This has required an investigation of low power hardware implementations for the wireless link and internal pad electronics, customized link protocols, mobile networking software and proxies which mediate between the mobility layer and existing applications. In addition, it was necessary to develop new applications which exploit the unique capabilities of the InfoPad (eg. pen and audio input without a keyboard), in order to demonstrate the full capabilities of the system solution being proposed. A presentation of the key results here will be made at the workshop.
It is clear that the most complete power consumption reduction possible is achieved by completely removing the function from the portable device. If it is assumed a high bandwidth communication link exists, then it is only necessary to leave computation in the pad that supports the communications and I/O functions. All applications, storage and support computation can be performed on fixed network attached resources. This also was found to have advantages in other areas as well, such as cost reduction (eg. low power and light weight mass storage is considerably more expensive than large server attached disks) and it simplifies the user support, since support is performed in centralized locations by knowledgeable system managers and not individually by the potentially vastly larger number of users of the portable units.
A number of important lessons and technologies have been learned and developed during the course of the InfoPad project research. It was hypothesized that a true system optimization should be attempted, which would range from the user interface, through the network and wireless link, to the terminal itself. Our results have clearly demonstrated that this was the correct approach in that the portable unit is significantly simpler than what could have been achieved by attempting to separately optimize individual segments of the overall system as is conventionally done. It was also clear, however, that even though we were willing to investigate custom solutions for much of our system design, it would have to be done in such a way that standard applications and network protocols could also be used in order to insure that full access would be available to existing network infrastructure and computational resources.
To provide this system optimization it was found necessary to divide the project into five efforts which would focus on an individual part of the problem and then to provide an organizational structure for interaction with the other groups in the system optimization process. The areas that were identified along with a short description were as follows:
It is clear in retrospect, however, that by actually having real hardware and actually making it useable resulted in the definition of a number of new research endeavors that are still being investigated. Probably most important of these is the low power design methodologies which have yielded results that we can now demonstrate on the InfoPad platform.