In this paper, we present design, implementation and evaluation of a novel predictive control framework to enable reliable distributed stream data processing, which features a Deep Recurrent Neural Network (DRNN) model for performance prediction, and dynamic grouping for flexible control. Specifically, we present a novel DRNN model, which makes accurate performance prediction with careful consideration for interference of co-located worker processes, according to multilevel runtime statistics. Moreover, we design a new grouping method, dynamic grouping, which can distribute/re-distribute data tuples to downstream tasks according to any given split ratio on the fly. So it can be used to re-direct data tuples to bypass misbehaving workers. We implemented the proposed framework based on a widely used Distributed Stream Data Processing System (DSDPS), Storm. For validation and performance evaluation, we developed two representative stream data processing applications: Windowed URL Count and Continuous Queries. Extensive experimental results show: 1) The proposed DRNN model outperforms widely used baseline solutions, ARIMA and SVR, in terms of prediction accuracy; 2) dynamic grouping works as expected; and 3) the proposed framework enhances reliability by offering minor performance degradation with misbehaving workers.

This paper presents a novel parallel platform, C-Storm (Copula-based Storm), for the computationally complex problem of fusion of heterogeneous data streams for inference. C-Storm is designed by marrying copula-based dependence modeling for highly accurate inference and a highly-regarded parallel computing platform Storm for fast stream data processing. C-Storm has the following desirable features: 1) C-Storm offers fast inference responses. 2) C-Storm provides high inference accuracies. 3) C-Storm is a general-purpose inference platform that can support data fusion applications. 4) C-Storm is easy to use and its users do not need to know deep knowledge of Storm or copula theory. We implemented C-Storm based on Apache Storm 1.0.2 and conducted extensive experiments using a typical data fusion application. Experimental results show that C-Storm offers a significant 4.7× speedup over a commonly used sequential baseline and higher degree of parallelism leads to better performance.

The Single Instruction Multiple Data (SIMD) architecture of Graphic Processing Units (GPUs) makes them perfect for parallel processing of big data. In this paper, we present the design, implementation and evaluation of G-Storm, a GPU-enabled parallel system based on Storm, which harnesses the massively parallel computing power of GPUs for high-throughput online stream data processing. G-Storm has the following desirable features: 1) G-Storm is designed to be a general data processing platform as Storm, which can handle various applications and data types. 2) G-Storm exposes GPUs to Storm applications while preserving its easy-to-use programming model. 3) G-Storm achieves high-throughput and low-overhead data processing with GPUs. 4) G-Storm accelerates data processing further by enabling Direct Data Transfer (DDT), between two executors that process data at a common GPU. We implemented G-Storm based on Storm 0.9.2 and tested it using three different applications, including continuous query, matrix multiplication and image resizing. Extensive experimental results show that 1) Compared to Storm, G-Storm achieves over 7χ improvement on throughput for continuous query, while maintaining reasonable average tuple processing time. It also leads to 2.3χ and 1.3χ throughput improvements on the other two applications, respectively. 2) DDT significantly reduces data processing time.

Automatic decision-making approaches, such as reinforcement learning (RL), have been applied to (partially) solve the resource allocation problem adaptively in the cloud computing system. However, a complete cloud resource allocation framework exhibits high dimensions in state and action spaces, which prohibit the usefulness of traditional RL techniques. In addition, high power consumption has become one of the critical concerns in design and control of cloud computing systems, which degrades system reliability and increases cooling cost. An effective dynamic power management (DPM) policy should minimize power consumption while maintaining performance degradation within an acceptable level. Thus, a joint virtual machine (VM) resource allocation and power management framework is critical to the overall cloud computing system. Moreover, novel solution framework is necessary to address the even higher dimensions in state and action spaces. In this paper, we propose a novel hierarchical framework for solving the overall resource allocation and power management problem in cloud computing systems. The proposed hierarchical framework comprises a global tier for VM resource allocation to the servers and a local tier for distributed power management of local servers. The emerging deep reinforcement learning (DRL) technique, which can deal with complicated control problems with large state space, is adopted to solve the global tier problem. Furthermore, an autoencoder and a novel weight sharing structure are adopted to handle the high-dimensional state space and accelerate the convergence speed. On the other hand, the local tier of distributed server power managements comprises an LSTM based workload predictor and a model-free RL based power manager, operating in a distributed manner. Experiment results using actual Google cluster traces show that our proposed hierarchical framework significantly saves the power consumption and energy usage than the baseline while achieving no severe latency degradation. Meanwhile, the proposed framework can achieve the best trade-off between latency and power/energy consumption in a server cluster.

In a distributed stream data processing system, an application is usually modeled using a directed graph, in which each vertex corresponds to a data source or a processing unit, and edges indicate data flow. In this paper, we propose a novel predictive scheduling framework to enable fast and distributed stream data processing, which features topology-aware modeling for performance prediction and predictive scheduling. For prediction, we present a topology-aware method to accurately predict the average tuple processing time of an application for a given scheduling solution, according to the topology of the application graph and runtime statistics. For scheduling, we present an effective algorithm to assign threads to machines under the guidance of prediction results. To validate and evaluate the proposed framework, we implemented it based on a highly-regarded distributed stream data processing platform, Storm, and tested it with 3 representative applications: word count (stream version), log stream processing and continuous query. Extensive experimental results show 1) The topology-aware prediction method offers an average accuracy of 84.2 percent. 2) The predictive scheduling framework reduces the average tuple processing time by 24.9 percent on average, compared to Storm's default scheduler.

The Single Instruction Multiple Data (SIMD) architecture of Graphic Processing Units (GPUs) makes them perfect for parallel processing of big data. In this paper, we present the design, implementation and evaluation of G-Storm, a GPU-enabled parallel system based on Storm, which harnesses the massively parallel computing power of GPUs for high-throughput online stream data processing. G-Storm has the following desirable features: 1) G-Storm is designed to be a general data processing platform as Storm, which can handle various applications and data types. 2) G-Storm exposes GPUs to Storm applications while preserving its easy-to-use programming model. 3) G-Storm achieves high-throughput and low-overhead data processing with GPUs. We implemented G-Storm based on Storm 0.9.2 and tested it using two different applications: continuous query and matrix multiplication. Extensive experimental results show that compared to Storm, G-Storm achieves over 7x improvement on throughput for continuous query, while maintaining reasonable average tuple processing time. It also leads to 2.3x throughput improvement for the matrix multiplication application.

In a distributed stream data processing system, an application is usually modeled using a directed graph, in which each vertex corresponds to a data source or a processing unit, and edges indicate data flow. In this paper, we propose a novel predictive scheduling framework to enable fast and distributed stream data processing, which features topology-aware performance prediction and predictive scheduling. For prediction, we present a topology-aware method to accurately predict the average tuple processing time of an application for a given scheduling solution, according to the topology of the application graph and runtime statistics. For scheduling, we present an effective algorithm to assign threads to machines under the guidance of prediction results. To validate and evaluate the proposed framework, we implemented it based on a highly-regarded distributed stream data processing platform, Storm, and tested it with two representative applications: word count (stream version) and log stream processing. Extensive experimental results show (1) The topology-aware prediction method offers an average accuracy of 83.7%. (2) The predictive scheduling framework reduces the average tuple processing time by 25.9% on average, compared to Storm's default scheduler.

In this paper, we study a Virtual Server Provisioning and Selection (VSPS) problem in distributed Data Centers (DCs) with the objective of minimizing the total operational cost while meeting the service response time requirement.We aim to develop general algorithms for the VSPS problem without assuming a particular queueing model for service processing in each DC. First, we present a Mixed Integer Linear Programming (MILP) formulation. Then we present a 3-step optimization framework, under which we develop a polynomial-time ln(N)-approximation algorithm (where N is the number of clients) along with a post-optimization procedure for performance improvement. We also show this problem is NP-hard to approximate and is not possible to obtain a better approximation ratio unless NP has TIME(nO(log log n)) deterministic time algorithms. In addition, we present an effective heuristic algorithm that jointly obtains the VS provisioning and selection solutions. Extensive simulation results are presented to justify effectiveness of the proposed algorithms.

In this paper, we present a novel cross-job framework for MapReduce scheduling, which aims to minimize the total processing time of a sequence of related jobs by combining reduce and map phases of two consecutive jobs and streaming data between them. The proposed framework has the following desirable properties: (1) It can accelerate the execution of a sequence of related MapReduce jobs by achieving a good tradeoff between data locality and parallelism. (2) It can support all the existing MapReduce applications with no changes to their source code. (3) It is a general framework, which can work with different scheduling algorithms. We built a new MapReduce runtime system called cross-job Hadoop by integrating the proposed cross-job framework into Hadoop. We conducted extensive experiments to evaluate its performance using PageRank and an Apache Pig application. Our experimental results show that the cross-job Hadoop can significantly reduce both the total processing time of a job sequence and the size of data transferred over the network.

Storm has emerged as a promising computation platform for stream data processing. In this paper, we first show inefficiencies of the current practice of Storm scheduling and challenges associated with applying traffic-aware online scheduling in Storm via experimental results and analysis. Motivated by our observations, we design and implement a new stream data processing system based on Storm, namely, T-Storm. Compared to Storm, T-Storm has the following desirable features: 1) based on runtime states, it accelerates data processing by leveraging effective traffic-aware scheduling for assigning/re-assigning tasks dynamically, which minimizes inter-node and inter-process traffic while ensuring no worker nodes are overloaded, 2) it enables fine-grained control over worker node consolidation such that T-Storm can achieve better performance with even fewer worker nodes, 3) it allows hot-swapping of scheduling algorithms and adjustment of scheduling parameters on the fly, and 4) it is transparent to Storm users (i.e., Storm applications can be ported to run on T-Storm without any changes). We conducted real experiments in a cluster using well-known data processing applications for performance evaluation. Extensive experimental results show that compared to Storm (with the default scheduler), T-Storm can achieve over 84% and 27% speedup on lightly and heavily loaded topologies respectively (in terms of average processing time) with 30% less number of worker nodes.

In this paper, we propose a service-aware approach to enhance survivability in virtualized data centers. The idea is to create and maintain a Survivable Virtual Infrastructure (SVI) for each service or tenant, which includes Virtual Machines (VMs) hosting the corresponding application and their backup VMs. A fundamental problem is to determine how to map each SVI to a data center network with minimum operational costs while satisfying each VM's resource requirements and bandwidth demands between VMs before and after failures. This problem can be naturally divided into two subproblems: VM Placement (VMP) and Virtual Link Mapping (VLM). We first present a general optimization framework. Then we propose an efficient algorithm for VMP, and a polynomial-time optimal algorithm for VLM, which can be used as subroutines in the framework. We also present an effective heuristic algorithm that jointly solves two subproblems. It has been shown by extensive simulation results based on the real VM workload traces collected from Syracuse University's green data center that compared to the First Fit Decreasing (FFD) and shortest path routing based baseline algorithm, the proposed algorithms significantly reduce the reserved bandwidth, and yield comparable results in terms of the number of active servers. \begin{keywords}Cloud Computing, Data Center, Service-aware, Survivability, Virtual Machine Management. \end{keywords}

In a virtualized data center, survivability can be enhanced by creating redundant Virtual Machines (VMs) as backup for VMs such that after VM or server failures, affected services can be quickly switched over to backup VMs. To enable flexible and efficient resource management, we propose to use a service-aware approach in which multiple correlated VMs and their backups are grouped together to form a Survivable Virtual Infrastructure (SVI) for a service or a tenant. A fundamental problem in such a system is to determine how to map each SVI to a physical data center network such that operational costs are minimized subject to the constraints that each VM's resource requirements are met and bandwidth demands between VMs can be guaranteed before and after failures. This problem can be naturally divided into two sub-problems: VM Placement(VMP) and Virtual Link Mapping (VLM). We present a general optimization framework for this mapping problem. Then we present an efficient algorithm for the VMP sub problem as well as a polynomial-time algorithm that optimally solves the VLM sub problem, which can be used as subroutines in the framework. We also present an effective heuristic algorithm that jointly solves the two sub problems. It has been shown by extensive simulation results based on the real VM data traces collected from the green data center at Syracuse University that compared with the First Fit Descending (FFD) and single shortest path based baseline algorithm, both our VMP+VLM algorithm and joint algorithm significantly reduce the reserved bandwidth, and yield comparable results in terms of the number of active servers.