2019 IEEE CLOUD COMPUTING

A Bi-layered Parallel Training Architecture for Large-Scale Convolutional Neural Networks

Abstract: Benefitting from large-scale training datasets and the complex training network, Convolutional Neural Networks (CNNs) are widely applied in various fields with high accuracy. However, the training process of CNNs is very time-consuming, where large amounts of training samples and iterative operations are required to obtain high-quality weight parameters. In this paper, we focus on the time-consuming training process of large-scale CNNs and propose a Bi-layered Parallel Training (BPT-CNN) architecture in distributed computing environments. BPT-CNN consists of two main components: (a) an outer-layer parallel training for multiple CNN sub networks on separate data subsets, and (b) an inner-layer parallel training for each sub network. In the outer-layer parallelism, we address critical issues of distributed and parallel computing, including data communication, synchronization, and workload balance. A heterogeneous aware Incremental Data Partitioning and Allocation (IDPA) strategy is proposed, where large-scale training datasets are partitioned and allocated to the computing nodes in batches according to their computing power. To minimize the synchronization waiting during the global weight update process, an Asynchronous Global Weight Update (AGWU) strategy is proposed. In the inner-layer parallelism, we further accelerate the training process for each CNN subnetwork on each computer, where computation steps of convolutional layer and the local weight training are parallelized based on task-parallelism. We introduce task decomposition and scheduling strategies with the objectives of thread-level load balancing and minimum waiting time for critical paths. Extensive experimental results indicate that the proposed BPT-CNN effectively improves the training performance of CNNs while maintaining the accuracy.

A  Cloud Brokerage Architecture for Efficient Cloud Service Selection

Abstract: The expanding cloud computing services offer great opportunities for consumers to find the best service and best pricing. Meanwhile, it also raises new challenges for consumers who need to select the best service out of such a huge pool since it will be time-consuming for consumers to collect the necessary information and analyze all service providers to make the decision. Therefore, in this paper, we propose a novel brokerage-based architecture in the cloud, where the cloud brokers is responsible for the service selection. We also design an efficient indexing structure, called B cloud-tree, for managing the information of a large number of cloud service providers. We then develop the service selection algorithm that recommends most suitable cloud services to the cloud consumers. We carry out extensive experimental studies on real and synthetic cloud data, and demonstrate a significant performance improvement over previous approaches.

A Cost-Efficient Router Architecture for HPC Inter-Connection Networks Design and Implementation

Abstract: High-radix routers with lower latency and higher bandwidth play an increasingly important role in constructing large-scale interconnection networks such as those used in super-computers and datacenters. The tile-based crossbar approach partitions a single large crossbar into many small tiles and can considerably reduce the complexity of arbitration while providing higher throughput than the conventional switch implementation. However, it is not scalable due to power consumption, placement, and routing problems. Inspired by non-saturated throughput theory, this paper proposes a scalable router microarchitecture, termed Multiport Binding Tile-based Router (MBTR). By aggregating multiple physical ports into a single tile a high-radix router can be flexibly organized into different tile arrays, thus the number of tiles and hardware overhead can be considerably reduced. For a radix-64 router MBTR achieves up to 50 ⇠ 75% reduction in memory consumption as well as wire area compared with a hierarchical switch. We theoretically deduce the sufficient and necessary conditions for the asymmetrical crossbar to achieve un-saturated relative 100% throughput. Based on this observation we analyze the MBTR throughput and derive the condition that should be satisfied by the MBTR design parameters to yield 100% throughput. We further discuss how to make a trade-off between MBTR parameters based on the constraints of performance, power and area. The simulation results demonstrate MBTR is indistinguishable from the YARC router in terms of throughput and delay, and can even outperform it by reducing potential contention for output ports. We have fabricated a 36-port MBTR chip at 28nm, providing 100 Gb/s bidirectional bandwidth per port, with a fall-through latency of just 30ns. Internally it runs at 9.6Tb/s, thus offering a speedup of 1.34⇥.

A Distributed Discovery of Communicating Resource Systems Models

Abstract: |In communicating resource systems (CRS), a set of independent, hierarchically composed resources communicate to realize  a business process. Examples of CRSs include REST ful Web services, cloud computing platforms and hierarchical distributed systems. Unfortunately, their complexity, distributed nature, and usually long lifespan can lead to problems like live locks, deadlock occurrences and invocation loops. This makes CRSs hard to maintain and manage; it also leads to a loss of QoS or system failures. Moreover, it is important to take into account, in system management, the exact way how the system gets to some desired or undesired state. Therefore, it is crucial to analyze processes, executed in a system to _nd and _x the above problems. Unfortunately, CRS process models are hardly ever available. Therefore, we present a new distributed algorithm, dRMA, discovering process models from CRS event logs. In order to reduce the representational bias between the discovered process model and the real process, we have de_ned a Communication Net, which can express features of CRSs: directed communication channels, hierarchy, and a resource perspective. Finally, we have evaluated dRMA on a cluster, and have showed that distributing jobs can signi_cantly reduce the time of process model discovery.

A Distributed Secure Outsourcing Scheme for Solving Linear Algebraic Equations in Ad Hoc Clouds

Abstract: The emerging ad hoc clouds form a new cloud computing paradigm by leveraging untapped local computation and storage resources. An important application of ad hoc clouds is to outsource computational intensive problems to nearby cloud agents. Specifically, for the problem of solving a linear algebraic equation (LAE), an outsourcing client assigns each cloud agent a sub problem, and then all involved agents apply a consensus-based algorithm to obtain the correct solution of the LAE in an iterative and distributed manner. However, such a distributed collaboration paradigm suffers from cyber security threats that undermine the confidentiality of the outsourced problem and the integrity of the returned results. In this paper, we identify a number of such security threats in this process, and propose a secure outsourcing scheme which not only preserves the privacy of the LAE parameters and the final solution from the participating agents, but also guarantees the correctness of the final solution. We prove that the proposed scheme has low computation complexity at each agent, and is robust against the identified security attacks. Numerical and simulation results are presented to demonstrate the effectiveness of the proposed method.

A Distributed Trust Evaluation Protocol with Privacy Protection for Intercloud

Abstract: Intercloud seeks to facilitate resource sharing among clouds. To support Intercloud, a trust evaluation framework among clouds and users is required. For trust evaluation, conventional protocols are typically based on a centralized  Architecture  focusing  on a one-way relationship. For Intercloud, the environment is highly dynamic and distributed, and relationships can be one-way or two-way (i.e., clouds provide services to each other). This paper presents a distributed trust evaluation protocol with privacy protection for Intercloud. The new contributions and innovative features are summarized below. First, feedback is protected by homomorphic encryption with verifiable secret sharing. Second, to cater to the dynamic nature of Intercloud, trust evaluation can be conducted in a distributed manner and is functional even when some of the parties are offline. Third, to facilitate customized trust evaluation, an innovative mechanism is used to store feedback, such that it can be processed flexibly while protecting feedback privacy. The protocol has been proved based on a formal security model. Simulations have been performed to demonstrate the effectiveness of the protocol. The results show that even when half of the clouds are malicious or offline, by choosing suitable operational parameters the protocol can still support effective trust evaluation with privacy protection.

A General Analysis Framework for Soft Real-Time Tasks

Abstract: Much recent work has been conducted on supporting soft real-time tasks on multiprocessors due to the multicore revolution. While most earlier works focus on the traditional sporadic task model with deterministic worst-case specification, several recent works investigate the stochastic nature of many workloads seen in practice, specifying task execution times using average-case provisioning instead of the worst case. Unfortunately, all the existing work on supporting soft real-time workloads ignores a simple practical fact that the job inter-arrival time (or task period) is also stochastic for many real-world applications. Adopting a fixed worst-case period to model all the arriving pattern is rather pessimistic and may result in significant capacity loss in practice. Based on these observations, we present a general soft real-time multiprocessor schedulability analysis framework in this paper for practical sporadic task systems specified by stochastic period and execution demand, following probability distributions. Our analysis can be generally applied to global tunable priority-based schedulers, which allow any job’s priority to be changed dynamically at runtime within a priority window of constant length. We have extensively evaluated the analysis framework using a MPEG video decoding case study and simulation-based experiments. Experimental results demonstrate significant advantages of our analysis, which yields over 200% and 50% improvements compared to existing analysis assuming worst-case task periods in terms of schedulability and magnitude of the derived tardiness bound, respectively.

A Hybrid Approach for Optimizing Carbon Footprint in InterCloud Environment

Abstract: This paper focuses on the problem of workload placement in an Inter Cloud with the view of minimizing the carbon footprint of such a computing environment. In order to reduce the ecological impact of the data center Greenhouse Gas (GhG) missions, this paper addresses the problem as a whole, by proposing a global mathematical formulation, based on the joint optimization of the Virtual Machine (VM) placement and their related traffics, along with a workload consolidation method and a cooling maximization technique that considers the dynamic behavior of the cooling fans. As the Virtual Machine Placement Problem (VMPP) is classified as an NP-hard problem, with the addition of the traffic embedding, the problem becomes more complex and stays NP-hard. Therefore, we propose a hybrid approach, for solving such problem and find good feasible solutions in a polynomial time. The results obtained from comparing with the exact method and other reference approaches help in assessing the efficiency of the proposed algorithm, as the carbon footprint costs are relatively close to the lower bound, with an average gap of about 3%, and found within a reasonable amount of time.

A Hybrid Approach for Optimizing Parallel Clustering Throughput using the GPU

Abstract: We introduce HYBRID-DBSCAN, that uses the GPU and CPUs for optimizing clustering throughput. The main idea is to exploit the memory bandwidth on the GPU for fast index searches, and optimize data transfers between host and GPU, to alleviate the potential negative performance impact of the PCIe interconnect. We propose and compare two GPU kernels that exploit grid-based indexing schemes to improve neighborhood search performance. We employ a batching scheme for host-GPU data transfers to obviate limited GPU memory, and exploit concurrent operations on the host and GPU. This scheme is robust with respect to both sparse and dense data distributions and avoids buffer overflows that would otherwise degrade performance. We evaluate our approaches on ionospheric total electron content datasets as well as intermediate-redshift galaxies from the Sloan Digital Sky Survey. HYBRID-DBSCAN outperforms the reference implementation across a range of application scenarios, including small workloads, which typically is the domain of CPU-only algorithms. We advance an empirical response time performance model of HYBRID-DBSCAN by utilizing the underlying properties of the datasets. With only a single execution of HYBRID-DBSCAN on a dataset, we are able to accurately predict the response time for a range of _ search distances.

A Modified Hierarchical Attribute-Based Encryption Access Control Method for Mobile Cloud Computing

Abstract: Cloud computing is an Internet-based computing pattern through which shared resources are provided to devices on demand. Its an emerging but promising paradigm to integrating mobile devices into cloud computing, and the integration performs in the cloud based hierarchical multi-user data-shared environment. With integrating into cloud computing, security issues such as data confidentiality and user authority may arise in the mobile cloud computing system, and it is concerned as the main constraints to the developments of mobile cloud computing. In order to provide safe and secure operation, a hierarchical access control method using modified hierarchical attribute-based encryption (M-HABE) and a modified three-layer structure is proposed in this paper. In a specific mobile cloud computing model, enormous data which may be from all kinds of mobile devices, such as smart phones, functioned phones and PDAs and so on can be controlled and monitored by the system, and the data can be sensitive to unauthorized third party and constraint to legal users as well. The novel scheme mainly focuses on the data processing, storing and accessing, which is designed to ensure the users with legal authorities to get corresponding classified data and to restrict illegal users and unauthorized legal users get access to the data, which makes it extremely suitable for the mobile cloud computing paradigms.

A Novel Task-Duplication Based Clustering Algorithm for Heterogeneous Computing Environments

Abstract: As a crucial task in heterogenous distributed systems, DAG-scheduling models a scheduling application with a set of distributed tasks by a Direct Acyclic Graph (DAG). The goal is to assign tasks to different processors so that the whole application can finish as soon as possible. Task Duplication-Based (TDB) scheme is an important technique addressing this problem. The main idea is to duplicate tasks on multiple machines so that the results of the duplicated tasks are available on multiple machines to trade computation time for communication time. Existing TDB algorithms enumerate and test all possible duplication candidates, and only keep the candidates that can improve the overall scheduling. We observe that while a duplication candidate is ineffective at the moment, after other duplications have been applied, this ineffective duplication candidate can become effective, which in turn can cause other ineffective duplications to become effective. We call this phenomenon the chain reaction of task duplication. We propose a novel Task Duplication based Clustering Algorithm (TDCA) to improve the schedule performance by utilizing duplication task more thoroughly. TDCA improves parameter calculation, task duplication, and task merging. The analysis and experiments are based on randomly generated graphs with various characteristics, including DAG depth and width, communication-computing cost ration, and variant computation power of processors. Our results demonstrate that the TDCA algorithm is very competitive. It improves the schedule makes pan of task duplication-based algorithms for heterogeneous systems for various communication computing costratios.