Zhihui Du

Robot as a Service in Cloud Computing

Service-oriented architecture and cloud computing are becoming a dominant computing paradigm, as all major computing companies are supporting this paradigm and more and more organizations are adopting this paradigm. Robotics and service-oriented robotics computing start to joint this new paradigm in the past five years and are now ready to participate in large scale. This paper reports our research on service-oriented robotics computing and our design, implementation, and evaluation of Robot as a Service (RaaS) unit. To fully qualify the RaaS as a cloud computing unit, we have kept our design to comply with the common service standards, development platforms, and execution infrastructure. We also keep the source code open and allow the community to configure the RaaS following the Web 2.0 principles of participation. Developers can add, remove, and modify the RaaS of their own. For this purpose, we have implemented our RaaS on Windows and Linux operating systems running on Atom and Core 2 Duo architectures. RaaS supports programming languages commonly used for service-oriented computing such as Java and C#. Special efforts have been made to support Microsoft Visual Programming Language (VPL) for graphic composition. We are working with high schools to use RaaS and VPL in robotics camps and robotics competitions.

Virtualization-based autonomic resource management for multi-tier Web applications in shared data center

As large data centers emerge, which host multiple Web applications, it is critical to isolate different application environments for security reasons and to provision shared resources effectively and efficiently to meet different service quality targets at minimum operational cost. To address this problem, we developed a novel architecture of resource management framework for multi-tier applications based on virtualization mechanisms. Key techniques presented in this paper include (1) establishment of the analytic performance model which employs probabilistic analysis and overload management to deal with non-equilibrium states; (2) a general formulation of the resource management problem which can be solved by incorporating both deterministic and stochastic optimizing algorithms; (3) deployment of virtual servers to partition resource at a much finer level; and (4) investigation of the impact of the failure rate to examine the effect of application isolation. Simulation experiments comparing three resource allocation schemes demonstrate the advantage of our dynamic approach in providing differentiated service qualities, preserving QoS levels in failure scenarios and also improving the overall performance while reducing the resource usage cost.

Typical Virtual Appliances: An optimized mechanism for virtual appliances provisioning and management

A computing infrastructure requirement in the cloud computing environment can be specified and composed using virtual appliances, which forms the infrastructure-as-a-service (IaaS). Due to the diversity of user requirements, a large number of virtual appliances may be needed. We propose a mechanism called Typical Virtual Appliances (TVAs), an efficient method for providing virtual appliances. In this paper, we present the concept of TVAs and formulate it as an optimization problem with given constraints. With analysis of the software download logs of real web sites, we discover that the number of user requirements follows a quadratic polynomial distribution, and the user requirements are clustered in nature. According to this finding, we develop a clustering-based TVAs generation algorithm, and we show that this algorithm can achieve the optimal result. The clustering algorithm can generate TVAs, which can be transformed to other virtual appliances easily and efficiently. We further design a TVA Management System (TVAMS) to support this mechanism. The simulation results show that our method can meet most of the user requirements efficiently with low storage overhead.

A tile-based parallel Viterbi algorithm for biological sequence alignment on GPU with CUDA

The Viterbi algorithm is the compute-intensive kernel in Hidden Markov Model (HMM) based sequence alignment applications. In this paper, we investigate extending several parallel methods, such as the wave-front and streaming methods for the Smith-Waterman algorithm, to achieve a significant speed-up on a GPU. The wave-front method can take advantage of the computing power of the GPU but it cannot handle long sequences because of the physical GPU memory limit. On the other hand, the streaming method can process long sequences but with increased overhead due to the increased data transmission between CPU and GPU. To further improve the performance on GPU, we propose a new tile-based parallel algorithm. We take advantage of the homological segments to divide long sequences into many short pieces and each piece pair (tile) can be fully held in the GPUs memory. By reorganizing the computational kernel of the Viterbi algorithm, the basic computing unit can be divided into two parts: independent and dependent parts. All of the independent parts are executed with a balanced load in an optimized coalesced memory-accessing manner, which significantly improves the Viterbi algorithms performance on GPU. The experimental results show that our new tile-based parallel Viterbi algorithm can outperform the wave-front and the streaming methods. Especially for the long sequence alignment problem, the best performance of tile-based algorithm is on average about an order magnitude faster than the serial Viterbi algorithm.

An adaptive model-free resource and power management approach for multi-tier cloud environments

With the development of cloud environments serving as a unified infrastructure, the resource management and energy consumption issues become more important in the operations of such systems. In this paper, we investigate adaptive model-free approaches for resource allocation and energy management under time-varying workloads and heterogeneous multi-tier applications. Specifically, we make use of measurable metrics, including throughput, rejection amount, queuing state, and so on, to design resource adjustment schemes and to make control decisions adaptively. The ultimate objective is to guarantee the summarized revenue of the resource provider while saving energy and operational costs. To validate the effectiveness, performance evaluation experiments are performed in a simulated environment, with realistic workloads considered. Results show that with the combination of long-term adaptation and short-term adaptation, the fluctuation of unpredictable workloads can be captured, and thus the total revenue can be preserved while balancing the power consumption as needed. Furthermore, the proposed approach can achieve better effect and efficiency than the model-based approaches in dealing with real-world workloads.

An autonomic provisioning framework for outsourcing data center based on virtual appliances

As outsourcing data centers emerge to host applications and services from many different organizations, it is critical for data center owners to isolate different applications while dynamically and optimally allocate sharable resources among them. To address this issue, we propose a virtual-appliance-based autonomic resource provisioning framework for large virtualized data centers. We present the architecture of the data center with enriched autonomic features. We define a non-linear constrained optimization model for dynamic resource provisioning and present a novel analytic solution. Key factors, including virtualization overhead and reconfiguration delay, are incorporated into the model. Experimental results based on a prototype demonstrate that the system-level performance has been greatly improved by taking advantage of fine-grained server consolidation, and the whole system exhibits flexible adaptation in failure scenarios. Experiments with the impact of switching delay also show the efficiency of the framework due to significantly reduced provisioning time.

A Waterfall Model to Achieve Energy Efficient Tasks Mapping for Large Scale GPU Clusters

High energy consumption has become a critical problem for supercomputer systems. GPU clusters are becoming an increasingly popular architecture for building supercomputers because of its great improvement in performance. In this paper, we first formulate the tasks mapping problem as a mini-mal energy consumption problem with deadline constraint. Its optimizing object is very different from the traditional mapping problem which often aims at minimizing make span or minimizing response time. Then a Waterfall Energy Consumption Model, which abstracts the energy consumption of one GPU cluster system into several levels from high to low, is proposed to achieve an energy efficient tasks mapping for large scale GPU clusters. Based on our Waterfall Model, a new task mapping algorithm is developed which tries to apply different energy saving strategies to keep the system remaining at lower energy levels. Our mapping algorithm adopts the Dynamic Voltage Scaling, Dynamic Resource Scaling and

An adaptive QoS management framework for VoD cloud service centers

beta

Controversy corner: Optimized QoS-aware replica placement heuristics and applications in astronomy data grid

-migration for GPU sub-task to significantly reduce the energy consumption and achieve a better load balance for GPU clusters. A task generator based on the real task traces is developed and the simulation results show that our mapping algorithm based on the Waterfall Model can reduce nearly 50% energy consumption compared with traditional approaches which can only run at a high energy level. Not only the task deadline can be satisfied, but also the task execution time of our mapping algorithm can be reduced.

Design and implementation of adaptive resource co-allocation approaches for cloud service environments

As cloud computing grows rapidly and Video-on-Demand (VoD) services becomes popular, it is critical and important to provide Quality of Service (QoS) to more customers under limited resources. To address this issue, we propose an adaptive QoS management framework for VoD cloud service centers. We present the architecture of the service center and then illustrate the QoS controlling process. To enhance the total revenue of the service provider, we define an optimization problem considering the charging model according to “pay-as-you-go” patterns. The QoS-aware Cache Replacement algorithm is then developed and described. Experiment results based on a prototype system and simulation tools demonstrate that the total revenue can be remarkably increased, because the QoS metrics of different classes of users could be guaranteed under varying workload and restricted resources.