The IP address consumption and the global routing table size are two of the vital parameters of the Internet growth. In this paper we quantitatively characterize the IPv4 address allocations made over the past six years and the global BGP routing table size changes during the same period of time. About 63,000 address blocks have been allocated since the beginning of the Internet, of which about 18,000 address blocks were allocated during our study period, from November 1997 to August 2004. Among these 18,000 allocations, 90\% of them started being announced into the BGP routing table within 75 days after the allocation, while 8\% of them has not been used up to now. %% LZ: is the follow 45% of allocated, or 45% of advertised prefixes? Among all the address blocks that have ever been used, 45\% of them were split into fragments smaller than the original allocated blocks; without these fragmentations, the current BGP table would have been about half of its current size. Furthermore, we found that the evolution of BGP routing table consists of both the appearance of new prefixes and the disappearance of old prefixes. While the change of the BGP routing table size only reflects the combined results of the two processes, the dynamics of either process is much higher than that of the BGP table size. %% LZ: this last sentence needs to be revised once we get the revised section 7 Finally, we classify routing prefixes into \emph{covering} and \emph{covered} ones, and examine their evolution separately. For the covered prefixes, which account for almost half of the BGP table size, we infer their practical motives such as multihoming, load balancing, and traffic engineering, etc., via a classification method.

Inspired by Zhang's work on flexible calibration technique, a new easy technique for calibrating a camera based on circular points is proposed. The proposed technique only requires the camera to observe a newly designed planar calibration pattern (referred to as the model plane hereinafter) which includes a circle and a pencil of lines passing through the circle's center, at a few (at least three) different unknown orientations, then all the five intrinsic parameters can be determined linearly. The main advantage of our new technique is that it needs to know neither any metric measurement on the model plane, nor the correspondences between points on the model plane and image ones, hence the whole calibration process becomes extremely simple. The proposed technique is particularly useful for those people who are not familiar with computer vision. Experiments with simulated data as well as with real images show that our new technique is robust and accurate. (C) 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.

Transport protocol for supporting multimedia streaming in mobile ad hoc networks has to cope with the rich dynamics, such as mobility-induced disconnection and reconnection, high out-of-order delivery ratios, channel errors, and network congestion. In this work, we design and implementate, ADTFRC, a TCP-friendly transport protocol for ad hoc networks. ADTFRC adapts wireline TFRC protocol to ad hoc networks with improved rate adaptation behavior, the capability of application layer framing and selective retransmission. ADTFRC detects different packet loss behaviors based on end-to-end measurements of multiple metrics. This allows ADTFRC to more accurately gauge the network behavior and achieve higher throughput. Simulations show that the performance of ADTFRC is higher than standard TFRC and TCP with explicit-link-failure-notification (ELFN) support in terms of throughput, rate adaptation behavior and application level quality while still maintaining the TCP-Friendliness property.

Large-scale sensor networks impose energy and communication constraints, thus it is difficult to collect data from each individual sensor node and process it at the sink. In this paper, we propose an efficient data-collection scheme that can be used for event monitoring or network-wide diagnosis. Our scheme relies on the well-known representation of data-contour maps, which trade off accuracy with the amount of samples. The scheme consists of three novel algorithms to build contour maps: distributed spatial and temporal data suppression, contour reconstruction at the sink via interpolation and smoothing, and an efficient mechanism to convey routing information over multiple hops. By reducing the number of transmissions required to convey relevant information to the sink, the proposed contour mapping scheme saves energy and improves network lifetime. In a sharp contrast to related work in this area, the scheme does not require all nodes to explicitly share information. The contour mapping scheme can be applied for tasks such as: (1) presenting a global picture of the network in both temporal and spatial domains, (2) being used as a diagnosis tool, e.g., to detect faulty sensors and to scan for residual energy, (3) working in concert with in-network aggregation schemes to further reduce the communication overhead of aggregation schemes. The proposed scheme imposes little processing and storage overhead, allowing for the sensor networking paradigm of 'dumb sensor, smart sink' which enables economical deployment of large-scale sensor networks. Simulation results show that our scheme is resilient to both high packet loss rate and measurement noise. The design is also energy efficient, resulting in up to an-order-of-magnitude power savings when compared with the base line scheme where every sensor sends its report to the sink. (c) 2005 Elsevier B.V. All rights reserved.

In this paper, we design a robust scheduling algorithm to ensure worst-case optimal performance when only coarse-grained channel state information (i.e., bounds on channel errors, but not the fine-grained error pattern) is available To solve this problem, we consider two coarse-grained channel error models and take a zero-sum game theoretic approach, in which the scheduler and the channel error act as non-cooperative adversaries in the scheduling process. Our results show that in the heavy channel error case, the optimal scheduler adopts a threshold form. It does not schedule a flow if the price (the flow is willing to pay) is too small, in order to maximize the system revenue. Among the scheduled flows, the scheduler schedules a flow with a probability inversely proportional to the flow price such that the risk of being caught by the channel error adversary is minimized. We also show that in the mild channel error model, the robust scheduling policy exhibits a balanced trade-off between a greedy decision and a conservative policy. The scheduler is likely to take a greedy decision if it evaluates the risk of encountering the channel error adversary now to be small. Therefore, robust scheduling does not always imply conservative decision. The scheduler is willing to take "risks" to expect higher gain in some scenarios. Our solution also shows that probabilistic scheduling may lead to higher worst-case performance compared to traditional deterministic policies. Finally, the current efforts show the feasibility to explore a probabilistic approach to cope with dynamic channel error conditions.