Fangzhe Chang

Topology inference in the presence of anonymous routers

Many topology discovery systems rely on traceroute to discover path information in public networks. However, for some routers, traceroute detects their existence but not their address; we term such routers anonymous routers. This paper considers the problem of inferring the network topology in the presence of anonymous routers. We illustrate how obvious approaches to handle anonymous routers lead to incomplete, inflated, or inaccurate topologies. We formalize the topology inference problem and show that producing both exact and approximate solutions is intractable. Two heuristics are proposed and evaluated through simulation. These heuristics have been used to infer the topology of the 6Bone, and could be incorporated into existing tools to infer more comprehensive and accurate topologies.

Optimal Resource Allocation in Clouds

Cloud platforms enable enterprises to lease computing power in the form of virtual machines. An important problem for such enterprise users is to understand how many and what kinds of virtual machines will be needed from clouds. We formulate demand for computing power and other resources as a resource allocation problem with multiplicity, where computations that have to be performed concurrently are represented as tasks and a later task can reuse resources released by an earlier task. We show that finding a minimized allocation is NP-complete. This paper presents an approximation algorithm with a proof of its approximation bound that can yield close to optimum solutions in polynomial time. Enterprise users can exploit the solution to reduce the leasing cost and amortize the administration overhead (e.g., setting up VPNs or configuring a cluster). Cloud providers may utilize the solution to share their resources among a larger number of users.

Topology discovery for public IPv6 networks

In just three decades the Internet has grown from a small experimental research network into a complex network of routers, switches, and hosts. Understanding the topology of such large scale networks is essential to the procurement of good architectural design decisions, particularly with respect to address allocation and distribution schemes.A number of techniques for IPv4 network topology already exist. Of these ICMP-based probing has shown to be most useful in determining router-level topologies of public networks. However, many of these techniques cannot be readily applied to IPv6 because of changes in the addressing scheme and ICMP behaviour. Furthermore, increases in the proliferation of equal-cost multi-path routing, and other forms of transient routing, indicate that traditional traceroute-based topology discovery approaches are becoming less effective in the Internet.This paper presents Atlas, a system that faciliates the automated capture of IPv6 network topology information from a single probing host. It describes the Atlas infrastructure and its data collection processes and discusses IPv6 network phenomena that must to be taken into account by the probing scheme. We also present some initial results from our probing of the 6Bone, currently the largest public IPv6 network. The results illustrate the effectiveness of the probing algorithm and also identify some trends in prefix allocation and routing policy.

Efficient video copy detection via aligning video signature time series

Various methods of content-based video copy detection have been proposed to find video copies in a large video database. In this paper, we represent video feature obtained by global and/or local detectors as signature time series. We observe that the curves of such time series under various kinds of modifications and transformations follow similar trends. Based on this observation, we propose to use linear segmentation to approximate the time series and extract major inclines from those linear segments. We develop a major incline-based fast alignment method to find potential alignment positions between the compared videos. Further, taking advantage of the major incline-based fast alignment, a Frame Insertion, Deletion, and Substitutions (FIDS) detection method is introduced to detect video copies in the presence of frame order changes. Our proposed solution is simple and generic. It can be combined with existing global or local feature descriptions, and with sequence or keyframe based matching schemes. It speeds up the video copy detection process by reducing the search space to the areas suggested by the potential alignments. Experiments using both the MUSCLE VCD 2007 and TRECVID CBCD 2009 datasets show that the proposed solution significantly accelerates the overall video copy detection process and at the same time achieves good precision.

Placement in Clouds for Application-Level Latency Requirements

CPU and device virtualization technology allows applications to be hosted on cloud platforms; some of the resulting benefits are lower cost and greater elasticity. In such cloud hosted applications, some components reside on the cloud while others, such as end users and components tied to physical devices, are located outside the cloud. Many applications, e.g., telecom services, have stringent latency requirements in terms of within how much time certain procedures must be completed. The application latency is strongly determined by the locations of all the interacting components that are both within and outside the cloud. In this paper, we study the problem of determining the optimal placement of the application components in the cloud so that the latency requirements of the application can be met. We present a precise formulation of the placement problem which includes a specification of the cloud platform, and collective latency expressions for application-level latency requirements. We show that Message Sequence Charts (MSCs), a widely-used mechanism for describing the execution of application procedures, can be naturally translated into our formalism of collective latency expressions. We present placement algorithms that exploit the Euclidean triangular inequality property of network topologies: (a) an exact algorithm for determining the most optimal placement but which has a worst-case exponential running time, and (b) an algorithm for determining a close to-optimal placement that has a fast polynomial running time. Additionally, we present an exact technique for partitioning a placement problem into smaller sub problems so that greater efficiency and accuracy can be achieved. We evaluate the performance of the algorithms on a representative telecom application --- a distributed deployment of the LTE Mobility Management Entity (MME). Our evaluation results show that our approximate algorithm can outperform a random placement by up to 49% for finding a successful placement.

Fast Near-Duplicate Video Retrieval via Motion Time Series Matching

This paper introduces a method for the efficient comparison and retrieval of near duplicates of a query video from a video database. The method generates video signatures from histograms of orientations of optical flow of feature points computed from uniformly sampled video frames concatenated over time to produce time series, which are then aligned and matched. Major incline matching, a data reduction and peak alignment method for time series, is adapted for faster performance. The resultant method is compact and robust against a number of common transformations including: flipping, cropping, picture-in-picture, photometric, addition of noise and other artifacts. We evaluate on the MUSCLE VCD 2007 dataset and a dataset derived from TRECVID 2009. Good precision (average 88.8%) at significantly higher speeds (average durations: 45 seconds for signature generation plus 92 seconds for a linear search of 81-second query video in a 300 hour dataset) than results reported in the literature are shown.

Authenticating Lossy Surveillance Video

Public camera feeds are increasingly being opened to use by multiple authorities (e.g., police, fire, traffic) as well as to the public. Because of the difficulty and insecurity of sharing cryptographic keys, these data are available in the clear. However, authorities must have a mechanism to assure trust in the video, that is, to authenticate it. While lossless video is straightforward to authenticate by cryptographic means, lossy video as may result from UDP, wireless, or transcoded transmissions, is more difficult to authenticate. We describe a method that combines a concise and efficiently computed video fingerprint with public key cryptography. Essential components of our approach are: the procedure to combine inexact video fingerprint with exact digital signature to enable lossy authentication, and matching for misaligned video via a major incline approach. Experimental results relate video fingerprint length to authentication accuracy and latency (time to authentication).

Optimal Resource Allocation for Batch Testing

Batch resource allocation problem arises in the context of executing a sequence of automated system tests or distributed computations where resources are pooled together and flexibly matched with requests. Minimizing resource allocation for a batch of processes reduces the resource management (e.g., setup) cost for the batch while allowing more users to share the resource pool simultaneously. The salient characteristic of the batch resource allocation problem is that while resources can be reused across different processes they are subject to mutually exclusive use for any individual process. We show that resource allocation for a single process can be solved in polynomial time whereas the general optimization problem is NP-complete. This motivates us to consider heuristics that can yield close to optimum solutions in polynomial time. We design several such heuristics and present their experimental comparison. Our experiments show that a technique based on a min-cost max-flow algorithm combined with ranked removal yields the best solution while having smallest running time.

Mitigating high latency outliers for cloud-based telecommunication services

Telecommunication applications are distinguished by their stringent requirements for availability and completion times. A highly available, low-latency, distributed data store is therefore a critical component of cloud-based realizations of telecommunication services. We present a systematic experimental evaluation of state-of-the-art database systems as components of telecommunication applications. We show that while their average latencies are well within the required time scales, the distribution of latencies exhibits a long tail of unacceptably large outliers which may significantly impair meeting the performance requirements of telecommunication applications. To address the observed phenomenon of high latency outliers, we present a new solution that is implemented in a Bell Labs system code named Flurry. Flurry is based on using the first response from a replica rather than waiting for all or a quorum of responses from replicas. To handle incorrect responses arising from message losses, Flurry uses a novel checking algorithm based on vector clocks to determine the correctness of a replicas response. We present experimental evaluation results which show that Flurry significantly reduces both the average response time and the probability of unacceptable response times to values that would allow meeting the availability and completion time thresholds required for telecommunication services.

Behavior based service composition

To enable flexible leveraging of the growing plethora of available web services, service clients should be automatically composed based on required behavior. In this paper, we present a foundational framework for behavior based composition. Services advertise their behavior as labeled transition systems with the action labels corresponding to their externally invocable operations. Query logics are defined in a simple extension of μ -calculus with modalities in which variables are allowed to occur. Query logics specify the desired behavior of the composition with the variables standing for the programs that need to be synthesized. We define a special subclass of programs, called deterministic and crash-free, which behave deterministically (even if the services used are non-deterministic) with all program steps successfully executing in whichever state the services may be in during entire execution. We present an algorithm that synthesises deterministic and crash-free programs whenever there exists such a solution. Since the μ -calculus is the most expressive logic for regular properties, our results yield a complete solution to the automatic composition problem for regular behavioral properties.