Gongxuan Zhang

Reliability and temperature constrained task scheduling for makespan minimization on heterogeneous multi-core platforms

Abstract We study the problem of scheduling tasks onto a heterogeneous multi-core processor platform for makespan minimization, where each cluster on the platform has a probability of failure governed by an exponential law and the processor platform has a thermal constraint specified by a peak temperature threshold. The goal of our work is to design algorithms that optimize makespan under the constraints of reliability and temperature. We first provide a mixed-integer linear programming (MILP) formulation for assigning and scheduling independent tasks with reliability and temperature constraints on the heterogeneous platform to minimize the makespan. However, MILP takes exponential time to finish. We then propose a two-stage heuristic that determines the assignment, replication, operating frequency, and execution order of tasks to minimize the makespan while satisfying the real-time, reliability, and temperature constraints based on the analysis of the effects of task assignment on makespan, reliability, and temperature. We finally carry out extensive simulation experiments to validate our proposed MILP formulation and two-stage heuristic. Simulation results demonstrate that the proposed MILP formulation can achieve the best performance in reducing makespan among all the methods used in the comparison. The results also show that the proposed two-stage heuristic has a close performance as the representative existing approach ESTS and a better performance when compared to the representative existing approach RBSA, in terms of reducing makespan. In addition, the proposed two-stage heuristic has the highest feasibility as compared to RBSA and ESTS.

Thermal-aware correlated two-level scheduling of real-time tasks with reduced processor energy on heterogeneous MPSoCs

Abstract With the exponential increase in power density and the relentless scaling of transistors in VLSI circuits over the past decades, modern high-performance processors fall into a predicament of high energy consumption and elevated chip temperature. Such increased energy consumption and chip temperature could induce significant economic, ecological, and technical problems. Thus, energy-efficient task scheduling with thermal consideration has become a pressing research issue in sustainable computing systems, especially for battery-powered real-time embedded systems with limited cooling techniques. This paper tackles the above challenge through scheduling tasks leveraging correlated optimizations at two different scales. Precisely, a two-level thermal-aware energy-efficient scheduling algorithm for real-time tasks on DVFS-enabled heterogeneous MPSoC systems is developed considering the constraints of task deadlines, task precedences, and chip peak temperature limit. At the processor level, a multi-processor model supporting dynamic voltage/frequency scaling is transformed to a virtual multi-processor model supporting only one fixed frequency level. At the core level, real-time tasks are assigned to individual cores of the virtual processor under the constraints of task precedence and peak temperature limit. Through nicely interleaving optimizations at both levels, high quality task scheduling solutions can be computed efficiently. Extensive simulations of synthetic real-time tasks and real-life benchmarks are performed to validate the proposed algorithm. Experimental results demonstrate the effectiveness of the proposed algorithm as compared to the benchmarking schemes.

An instances placement algorithm based on disk I/O load for big data in private cloud

In generally, large companies or organizations have the demand of big data processing and they do not want to entrust their business processes and data to third parties (Amazon, Google, etc.). The private cloud could meet their needs. In private cloud, tasks run at multiple instances (also known as virtual machines), which could be paced in different physical nodes. Obviously, the instances which be used to process big data need higher CPU and disk performance than other kinds of instances. If the instances of disk resource consuming are placed in the same physical node, clearly, the disk I/O bandwidth would be used up quickly that would affect the performance of the entire node seriously. This paper proposes an instances placement algorithm FFDL that based on disk I/O for private cloud environment to deal with big data that would adopt the disk I/O load balancing strategy and reduce competition for the disk I/O load between instances. We have validated our approach by conducting a performance evaluation study on the open source private cloud platform-Openstack. The results demonstrate that our algorithm has immense potential as it offers significant computation time savings than the Greedy algorithm and demonstrates high potential for the improvement of disk I/O load balancing in the entire private cloud system for the big data.

Snore related signals processing in a private cloud computing system

Snore related signals (SRS) have been demonstrated to carry important information about the obstruction site and degree in the upper airway of Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS) patients in recent years. To make this acoustic signal analysis method more accurate and robust, big SRS data processing is inevitable. As an emerging concept and technology, cloud computing has motivated numerous researchers and engineers to exploit applications both in academic and industry field, which could have an ability to implement a huge blue print in biomedical engineering. Considering the security and transferring requirement of biomedical data, we designed a system based on private cloud computing to process SRS. Then we set the comparable experiments of processing a 5-hour audio recording of an OSAHS patient by a personal computer, a server and a private cloud computing system to demonstrate the efficiency of the infrastructure we proposed.

Briareus: Accelerating Python Applications with Cloud

Briareus provides convenient tools to make use of computing resources provided by cloud to accelerate Python applications. In this paper, three techniques are presented. First, some of the functions in a Python program can be migrated to cloud and be evaluated using the hardware and software provided by that cloud platform, while the other parts still running locally. Second, Briareus can automatically parallelize specified loops in a program to accelerate it. And third, specified functions can be called asynchronously after being patched, so that two or more functions can be evaluated simultaneously. By combining these three methods, a Python application can make part of itself to run in a remote cloud platform in parallel. To use Briareus, developers do not need to modify the existing source much, but only need to insert some descriptive comments and invoke a patching function at the beginning. Experiments show that Briareus can significantly speed up the running of programs written by Python, especially for those for scientific and engineering computing. The early beta version of Briareus has been developed for testing and all sources are accessible to public via GitHub and installable via PyPI.

A New Path-Constrained Rendezvous Planning Approach for Large-Scale Event-Driven Wireless Sensor Networks

We study the problem of employing a mobile-sink into a large-scale Event-Driven Wireless Sensor Networks (EWSNs) for the purpose of data harvesting from sensor-nodes. Generally, this employment improves the main weakness of WSNs that is about energy-consumption in battery-driven sensor-nodes. The main motivation of our work is to address challenges which are related to a network’s topology by adopting a mobile-sink that moves in a predefined trajectory in the environment. Since, in this fashion, it is not possible to gather data from sensor-nodes individually, we adopt the approach of defining some of the sensor-nodes as Rendezvous Points (RPs) in the network. We argue that RP-planning in this case is a tradeoff between minimizing the number of RPs while decreasing the number of hops for a sensor-node that needs data transformation to the related RP which leads to minimizing average energy consumption in the network. We address the problem by formulating the challenges and expectations as a Mixed Integer Linear Programming (MILP). Henceforth, by proving the NP-hardness of the problem, we propose three effective and distributed heuristics for RP-planning, identifying sojourn locations, and constructing routing trees. Finally, experimental results prove the effectiveness of our approach.

A New Architecture in Greenhouse Soil Solution Monitoring System Based on ZigBee Protocol

Nowadays in the greenhouse monitoring system, there are some shortages: a. the concentration of nutrients (such as N, P plasma) in soil solution as a greenhouse environment parameter is often overlooked; b. small ZigBee system can’t meet the needs of real-time data storage and remote monitoring. Basing on the situation, this paper designs a new architecture in greenhouse soil solution monitoring system based on ZigBee protocol. The architecture has several small sub-networks. Each of these subnets forms an Ad-hoc network by ZigBee protocol. And nodes in the sub-network can real-time monitor the soil solution concentrations of greenhouse and send data to the coordinator after carrying out data cleaning. The coordinator will send real-time data to a PC monitoring machine and the Cloud. Using PC monitoring machine staff can monitor greenhouse crop conditions. And by utilizing mobile devices access to the Cloud, customers can also monitor soil solution concentration. The experimental results show that the system is stable, real-time and accurate

Genetic algorithm for energy-efficient clustering and routing in wireless sensor networks

Abstract Wireless sensor networks have been employed widely in various fields, including military, health care, and manufacturing applications. However, the sensor nodes are limited in terms of their energy supply, storage capability, and computational power. Thus, in order to improve the energy efficiency and prolong the network life cycle, we present a genetic algorithm-based energy-efficient clustering and routing approach GECR. We add the optimal solution obtained in the previous network round to the initial population for the current round, thereby improving the search efficiency. In addition, the clustering and routing scheme are combined into a single chromosome to calculate the total energy consumption. We construct the fitness function directly based on the total energy consumption thereby improving the energy efficiency. Moreover, load balancing is considered when constructing the fitness function. Thus, the energy consumption among the nodes can be balanced. The experimental results demonstrated that the GECR performed better than other five methods. The GECR achieved the best load balancing with the lowest variances in the loads on the cluster heads under different scenarios. In addition, the GECR was the most energy-efficient with the lowest average energy consumed by the cluster heads and the lowest energy consumed by all the nodes.

Certificateless Public Verification for the Outsourced Data Integrity in Cloud Storage

By advances in cloud storage systems, users have access to the data saved in the cloud and can manipulate the data without limitation of time and place. As the data owner no longer possesses data physically, he is required to ensure the integrity of the data stored in the cloud with the public key given by public key infrastructure (PKI). Thus the security of PKI and certificates are essential. However, there are numerous security risks in the traditional PKI and it is complex to administer the certificates. Certificateless public key cryptography is used in this paper to solve these problems. We also use elliptic curve group to reduce computation overhead. In this paper, we design a certificateless public verification mechanism to check the integrity of data outsourced in the cloud and we further extend it to support a multiuser group by batch verification. Specifically, a public verifier who replaces the data owner to check the integrity in the proposed scheme does not require to manage any certificates...

A Fast Algorithm of Compressed Sensing for 2D Signals

ABSTRACT Compressed sensing (CS) is a new signal acquisition method that can do sampling and compression of signals simultaneously. In order to reduce the signal reconstruction time of CS algorithms and lower the growth rate of the reconstruction time when increasing the size of signals, this paper proposes the algorithm of block whole orthogonal matching pursuit (BWOMP), which is a fast CS algorithm based on the method of orthogonal matching pursuit (OMP) for two-dimension (2D) signals. BWOMP defines a measurement parameter named whole-correlation. At each iteration, instead of computing the correlation between each atom and 1D residuals, the whole-correlation is computed as the correlation between the atom and the 2D residuals. After that, an approximation of the 2D signal is generated directly by BWOMP. By reducing the number of the iterations, this method can significantly lower the computational complexity. On the other hand, BWOMP introduces the concept of block compressed sensing (BCS), and redesigns the block size and the observation matrix. BCS reduces the consumption of computational resources (i.e. memory and CPU cycles) by reducing the size of variables (especially the matrixes). The experimental comparisons show that, in comparison with OMP, BWOMP can save at least 80% reconstruction time, which makes the increasing rate of reconstruction time linear. The results indicate that the proposed algorithm may have great performance advantage for complex cases.