- Advanced Data Storage Technologies
- Caching and Content Delivery
- Parallel Computing and Optimization Techniques
- Cloud Computing and Resource Management
- Distributed and Parallel Computing Systems
- Reliability and Maintenance Optimization
- Distributed systems and fault tolerance
- Statistical Distribution Estimation and Applications
- IoT and Edge/Fog Computing
- Peer-to-Peer Network Technologies
- Life Cycle Costing Analysis
- Cloud Data Security Solutions
- Security and Verification in Computing
- Scientific Computing and Data Management
- Optical Network Technologies
- Semiconductor Lasers and Optical Devices
- Data Mining Algorithms and Applications
- Cryptography and Data Security
- Advanced Battery Technologies Research
- Advanced Photonic Communication Systems
- Advanced ceramic materials synthesis
- Photonic Crystal and Fiber Optics
- Customer churn and segmentation
- Bayesian Modeling and Causal Inference
- Power Line Inspection Robots
ShanghaiTech University
2017-2024
Beijing Institute of Technology
2014-2022
Corning (United States)
2019-2021
Shanghai Institute of Microsystem and Information Technology
2020
University of Chinese Academy of Sciences
2020
Nanjing University of Posts and Telecommunications
2019
New York Institute of Technology
2019
Research & Development Corporation
2019
Centre for High Performance Computing
2019
Hunan University
2012-2016
MapReduce has become an important distributed processing model for large-scale data-intensive applications like data mining and web indexing. Hadoop-an open-source implementation of is widely used short jobs requiring low response time. The current Hadoop assumes that computing nodes in a cluster are homogeneous nature. Data locality not been taken into account launching speculative map tasks, because it assumed most maps data-local. Unfortunately, both the homogeneity assumptions satisfied...
Energy conservation is a major concern in cloud computing systems because it can bring several important benefits such as reducing operating costs, increasing system reliability, and prompting environmental protection. Meanwhile, power-aware scheduling approach promising way to achieve that goal. At the same time, many real-time applications, e.g., signal processing, scientific have been deployed clouds. Unfortunately, existing energy-aware algorithms developed for clouds are not task...
Augmented reality (AR) applications that overlay the perception of real world with digitally generated information are on cusp commercial viability. AR has appeared in several platforms like Microsoft HoloLens and smartphones. They extend user experience beyond two dimensions supplement normal 3D a user. A typical location-based multi-player application works through three-step process, wherein system collects sensory data from world, identifies objects based their context, finally, renders...
Highly sensitive and selective gas sensors hold great potential in disease diagnosis. However, the rational design of non-noble-metal-based, high-efficiency sensing materials for trace detection remains a crucial challenge. Herein, chemiresistive sensor that can detect parts per billion (ppb)-level acetone was realized based on three-dimensional (3D) α-Fe2O3/ZnO nanocages, which were achieved by simple encapsulation calcination process. In particular, we found Fe species play an intriguing...
This work presents a method to 1) statistically characterize the complex geometry of porous material microstructures, 2) parameterize these random microstructures into condense, feature rich metrics, and 3) relate metrics properties such as permeability. is applied synthetic cordierite aluminum titanate materials covering range 40%–70% porosity 8–18 μm median pore size (as measured by mercury intrusion porosimetry), using X-ray computed tomography images microstructure. Direct simulations...
Abstract For a component or system subject to stochastic degradation with sporadic jumps that occur at random times and have sizes, we propose model the cumulative using single process based on characteristics of Lévy subordinators, class nondecreasing processes. Based inverse Fourier transform, derive new closed‐form reliability function probability density for lifetime, represented by measures. The derived traditional convolution approach common models such as gamma jumps, is revealed be...
Running evolutionary algorithms in parallel is an intuitive way to speed up the process of solving large-scale multi-objective optimization problems, which have hundreds or thousands decision variables. However, framework existing seriously limits their parallelization. During each iteration, environmental selection operators present need collect and compare all candidate solutions balance convergence diversity, thus dividing whole into a series dependent sub-processes resulting frequent...
Load balancing for clusters has been investigated extensively, mainly focusing on the effective usage of global CPU and memory resources. However, previous CPU- or memory-centric load schemes suffer significant performance drop under I/O-intensive workloads due to imbalance I/O load. To solve this problem, we propose two simple yet I/O-aware load-balancing types clusters: (1) homogeneous where nodes are identical (2) heterogeneous clusters, which comprised a variety with different...
We use Lévy subordinators and non-Gaussian Ornstein–Uhlenbeck processes to model the evolution of degradation with random jumps. The superiority our models stems from flexibility such in modeling stylized features data series as jumps, linearity/nonlinearity, symmetry/asymmetry, light/heavy tails. Based on corresponding Fokker–Planck equations, we derive explicit results for reliability function lifetime moments terms Laplace transforms, represented by measures. Numerical experiments are...
Cluster computing has emerged as a primary and cost-effective platform for running parallel applications, including communication-intensive applications that transfer large amount of data among the nodes cluster via interconnection network. Conventional load balancers have proven effective in increasing utilization CPU, memory, disk I/O resources cluster. However, most existing load-balancing schemes ignore network resources, leaving an opportunity to improve bandwidth networks on clusters...
A critical problem with parallel I/O systems is the fact that disks consume a significant amount of energy. To design economically attractive and environmentally friendly systems, we propose an energy-aware prefetching strategy (PRE-BUD) for disk buffers. We introduce new architecture provides energy savings using buffer while maintaining high performance. There are two configurations: (1) adding extra to accommodate prefetched data, (2) utilizing existing as disk. PRE-BUD not only able...
Parallel disk systems consume a significant amount of energy due to the large number disks. To design economically attractive and environmentally friendly parallel systems, in this paper we evaluate an energy-aware prefetching strategy for consisting small buffer disks data Using temporarily handle requests disks, can keep low-power mode as long possible. Our algorithm aims group many idle periods form periods, which turn allow remain standby state save energy. achieve goal, utilize...
In the past decade, parallel disk systems have been developed to address problem of I/O performance. A critical challenge with modern is that disks consume a significant amount energy in servers and high performance computers. To conserve consumption systems, one can immediately spin down when are idle; however, spinning might not be able produce savings due penalties operations. Unlike powering up CPUs, need physical movements. Therefore, provided by operations must offset substantially...
An increasing number of commodity clusters are connected to each other by public networks, which have become a potential threat security sensitive parallel applications running on the clusters. To address this issue, we developed Message Passing Interface (MPI) implementation preserve confidentiality messages communicated among nodes in an unsecured network. We focus M PI rather than protocols, because is one most popular communication protocols for computing Our MPI implementation-called...
Recent studies indicate that the energy cost and carbon footprint of data centers have become exorbitant. It is a demanding challenging task to reduce consumption in large-scale storage systems modern centers. Most conservation techniques inevitably adverse impacts on parallel disk systems. To address reliability issues energy-efficient disks, we propose reliable RAID system called REED, which aims at improving both efficiency by seamlessly integrating HDDs SSDs. At heart REED...
Many energy conservation techniques have been proposed to achieve high efficiency in disk systems. Unfortunately, growing evidence shows that energy-saving schemes drives usually negative impacts on storage Existing reliability models are inadequate estimate of parallel systems equipped with techniques. To solve this problem, we propose a mathematical model - called MINT evaluate the system where mechanisms implemented. In paper, focus modeling two well-known Popular Disk Concentration...
We develop a mathematical model - MREED to quantitatively evaluate the failure rate of energy-efficient parallel storage systems. The Power-Aware Redundant Array Inexpensive Disk (PARAID) aims reduce energy use commodity server-class disks without specialized hardware. goal PARAID is skewed striping pattern adapt system load by changing number powered disks. By spinning down during light workloads, can power consumption, while still meeting performance demands. show that be used estimate...
This paper addresses an issue of erasure-coded data archival, where (k + r; k) erasure codes are employed to archive rarely accessed replicas. The traditional synchronous encodingprocess neither leverages the existence replicas, nor handles encoding operations in a decentralized manner. To overcome these drawbacks, we exploit pipelined processes boost archival performance on storage clusters. First, propose two layouts called [D P] <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...
The analysis of multiple dependent degradation processes is a challenging research problem in the reliability field, especially for complex with random jumps. To integrally handle jumps uncertainties and dependence among processes, we construct general multi-dimensional Lévy to describe engineering systems. evolution each process can be modeled by one-dimensional subordinator marginal measure. all dimensions described copulas associated multiple-dimensional measure obtained from measures...