There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Gctf: Real-time CTF determination and correction

This paper presents control and coordination algorithms for groups of vehicles. The focus is on autonomous vehicle networks performing distributed sensing tasks where each vehicle plays the role of a mobile tunable sensor. The paper proposes gradient descent algorithms for a class of utility functions which encode optimal coverage and sensing policies. The resulting closed-loop behavior is adaptive, distributed, asynchronous, and verifiably correct.

Coverage control for mobile sensing networks

/pdf/a-break-in-the-clouds-towards-a-cloud-definition-svovsw1r5a.pdf

A Break in the Clouds: Towards a Cloud Definition

The increasing complexity of computer systems and communication networks induces tremendous requirements for trust and security. This special issue includes topics on trusted computing, risk and reputation management, network security and survivable computer systems/networks. These issues have evolved into an active and important area of research and development. The past decade has witnessed a proliferation of concurrency and computation systems for practice of highly trust, security and privacy, which has become a key subject in determining future research and development activities in many academic and industrial branches. This special issue aims to present and discuss advances of current research and development in all aspects of trusted computing and network security. In addition, this special issue provides snapshots of contemporary academia work in the field of network trusted computing. We prepared and organized this special issue to record state-of-the-art research, novel development and trends for future insight in this domain. In this special issue, 14 papers have been accepted for publication, which demonstrate novel and original work in this field. A detailed overview of the selected works is given below.

Frontier technologies of trust computing and network security

Wavelength Division Multiplexing (WDM) is a promising technology to utilize the huge bandwidth of optical fiber. Wavelength-routed all-optical network is a suitable architecture to serve as the backbone of wide-area networks by providing lightpath services. Blocking has been the key performance index in the design of a wavelength-routed network. Existing research demonstrates that an effective routing and wavelength assignment (RWA) algorithm and wavelength conversion are the two primary vehicles for improving the blocking performance. However, these two issues have largely been investigated separately in that the existing RWA algorithms have seldom considered the presence of wavelength conversion, while the wavelength converter placement algorithms have largely assumed that a static routing and random wavelength assignment algorithm is employed. 
In this dissertation, we first investigate the wavelength converter placement problem for different RWA algorithms. Under the fixed-alternate routing (FAR) algorithm, we propose a heuristic algorithm called Minimum Blocking Probability First (MBPF) for wavelength converter placement. Under the least-loaded routing (LLR) algorithm, we propose another heuristic algorithm called Weighted Maximum Segment Length (WMSL). We observe that the proposed algorithms not only outperform existing wavelength converter placement algorithms by a large margin, but they also can achieve almost the same performance comparing with full wavelength conversion under the same RWA algorithm. The second work is to investigate the dynamic routing algorithm in the presence of wavelength conversion. We show that existing dynamic routing algorithms cannot utilize the fiber link resources efficiently in the presence of wavelength conversion. We then propose a weighted least-congestion routing (WLCR) algorithm that considers both the distribution of free wavelengths and the lengths of each route jointly. Finally, we propose the sparse-partial wavelength conversion architecture, which can save the number of wavelength converters significantly. In this architecture, two problems have been examined. The first one is the wavelength assignment problem, and the second one is the converter placement problem.

Rwa and wavelength conversion in wavelength-routed all-optical wdm networks

It has become a common practice to train large machine learning (ML) models across a cluster of computing nodes connected by RDMA-enabled networks. However, the communication overhead caused by parameter synchronization deteriorates the performance of such distributed ML (DML), especially in a large-scale setting. This paper tackles this issue by developing a traffic management scheme to support DML traffic, called TMDML (Traffic Management for DML), which needs only a minor modification to the existing RDMA congestion control scheme DCQCN. We assume that there is only one instance of DML workload running in a network. Existing literature has shown that Fat-Tree, a predominant topology in the data center, poorly supports DML compared with BCube. With our proposed TMDML, training DML in Fat-Tree can achieve better performance than that in BCube. We first study the impact of multi-bottlenecks on DML via NS-3-based simulations. The results show that DCQCN is inefficient for DML traffic in the multi-bottlenecks scenario. To mitigate the impact of multi-bottlenecks, we propose an optimization model to minimize the maximum flow completion time (FCT) while stabilizing the queues, and then apply the Lyapunov optimization technique to solve the problem. For all the practical purposes, we present two heuristic implementations of TMDML for different deployment requirements. We evaluate the performance of our proposals by simulation, comparing it with DCQCN. We use All-Reduce parameter synchronization in Fat-Tree and BCube with traffic trace of modern deep neural network models, including AlexNet, ResNet50, and VGG-16. Our proposals can achieve up to 59% of the time reduction.

Traffic Management for Distributed Machine Learning in RDMA-enabled Data Center Networks

Fingerprint is one of the representative methods for wireless indoor localization. It uses a fingerprint database (measured in the offline phase) and the current received signal strengths (RSSs) (measured by the user's device in the online phase) to determine the location of this device. However, the RSSs and hence the localization accuracy would be affected by time-varying environmental factors (e.g., number of people in a shopping mall). In this paper, we propose a new method for wireless localization in time-varying indoor environments. In the offline phase, the proposed method measures extra information: it measures $E$ fingerprint databases for $E$ respective environmental conditions, where $E$ is a design parameter (e.g., $E=2$ for the peak period and the non-peak period in a shopping mall). In the online phase, it leverages the extra information for better localization in time-varying indoor environment, even when the current environmental condition is different from the ones considered in the offline phase. The proposed method is particularly suitable for the indoor venues for which their primary concern is to provide good quality localization services while they could afford a moderate amount of extra resources for one-off measurement in the offline phase (e.g., exhibition centers, airports, shopping malls, etc.). We conduct a simulation experiment and a real-world experiment to demonstrate that the proposed method gives accurate localization.

Multi-Fingerprint for Wireless Localization in Time-Varying Indoor Environment

Calibrators and sniffers have been used in the literature to proactively update the functional relationship between the received signal strength and the distance for wireless localization in time-varying indoor venue, where calibrators and sniffers are Wi-Fi transmitters and Wi-Fi receivers respectively. To be effective, these devices should be suitably placed in the indoor venue. Let there be N calibrators and M sniffers, d i,j be the distance between calibrator i and sniffer j, and RSS i,j be the received signal strength measured by sniffer j from calibrator i. The points (d 1,1 , RSS 1,1 ), (d 1,2 , RSS 1,2 ), …, (d N,M , RSS N,M ) are used to estimate the functional relationship between the received signal strength and the distance. It is desirable that d 1,1 , d 1,2 , …, d N,M are uniformly scattered so that the estimated functional relationship is more accurate for better localization. In this paper, we propose to minimize the discrepancy of d 1,1 , d 1,2 , …, d N,M in order to determine the optimal locations of the N calibrators and the M sniffers. We formulate this new problem (named minimal discrepancy placement problem) and design an efficient optimization method to solve it. We conduct simulation and real-world experiments to demonstrate that minimal discrepancy placement can effectively improve localization accuracy.

Xiaowen Chu

Papers

Frontier technologies of trust computing and network security

Rwa and wavelength conversion in wavelength-routed all-optical wdm networks

Traffic Management for Distributed Machine Learning in RDMA-enabled Data Center Networks

Multi-Fingerprint for Wireless Localization in Time-Varying Indoor Environment

Minimal Discrepancy Placement of Sniffers and Calibrators for Wireless Indoor Localization