Search

Deep neural networks bear substantial cloud computational loads and often surpass client devices' capabilities. Research has concentrated on reducing the inference burden of convolutional neural networks processing images. Unstructured pruning, which leads to sparse matrices requiring specialized hardware, has been extensively studied. However, neural networks trained with tabular data and structured pruning, which produces dense matrices handled by standard hardware, are less explored. We compare two approaches: 1) Removing neurons followed by training from scratch, and 2) Structured pruning followed by fine-tuning through additional training over a limited number of epochs. We evaluate these approaches using three models of varying sizes (1.5, 9.2, and 118.7 million parameters) from Kaggle-winning neural networks trained with tabular data. Approach 1 consistently outperformed Approach 2 in predictive performance. The models from Approach 1 had 52%, 8%, and 12% fewer parameters than the original models, with latency reductions of 18%, 5%, and 5%, respectively. Approach 2 required at least one epoch of fine-tuning for recovering predictive performance, with further fine-tuning offering diminishing returns. Approach 1 yields lighter models for retraining in the presence of concept drift and avoids shifting computational load from inference to training, which is inherent in Approach 2. However, Approach 2 can be used to pinpoint the layers that have the least impact on the model's predictive performance when neurons are removed. We found that the feed-forward component of the transformer architecture used in large language models is a promising target for neuron removal.

Cloud computing and deep learning, the recent trends in the software industry, have enabled small companies to scale their business up rapidly. However, this growth is not without a cost – deep learning models are related to the heaviest workloads in cloud data centers. When the business grows, the monetary cost of deep learning in the cloud also grows fast. Deep learning practitioners should be prepared and equipped to limit the growing cost. We emphasize monetary cost instead of computational cost although often the same methods decrease both types of cost. We performed a systematic literature review on the methods to control the cost of deep learning. Our library search resulted in 16,066 papers from three article databases, IEEE Xplore, ACM Digital Library, and Scopus. We narrowed them down to 112 papers that we categorized and summarized. We found that: 1) Optimizing inference has raised more interest than optimizing training. Widely used deep learning libraries already support inference optimization methods, such as quantization, pruning, and teacher-student. 2) The research has been centered around image inputs, and there seems to be a research gap for other types of inputs. 3) The research has been hardware-oriented, and the most typical approach to control the cost of deep learning is based on algorithm-hardware co-design. 4) Offloading some of the processing to client devices is gaining interest and can potentially reduce the monetary cost of deep learning.

In IIoT, WirelessHART is one of the reliable wireless network standards since it can improve the productivity and efficiency. Recently, in Industrial area, the enterprise and the researcher have been paid attention to a mobile device, which will become one of the key points in the industry due to its potential ability. However, since WirelessHART existing researches handle with a mobile device as a single point-to-point viewpoint, they will happen a disconnection and a wasteful communication resource problem when they use a mobile device. In order to solve this limitation, we propose a proactive and cost-effective scheme for the mobility by focusing on a mobility as a group viewpoint. By using onion-peeling method, the proposed scheme can select some fixed devices to communicate with a mobile device and prevent redundant communication provision for a mobile device. The simulation results show that the proposed scheme is more cost-efficient and reliable when it compares with the existing researches of WirelessHART.

Despite the potential benefits of Internet of Underwater Things, a number of issues hinder its realization, including the need for communication reliability and cost-effectiveness. This article aims to optimize network design to implement cost-effective underwater acoustic sensor networks (UASNs) with 3D topology while supporting diverse communication quality of service (QoS) requirements. First, we present an analytical framework based on a queueing system that evaluates communication performances of UASNs, wherein each underwater sensor distributed within a 3D space under the sea surface performs fountain code (FC)-based automatic repeat request (ARQ) transmissions under the slotted-Aloha medium access control protocol. Under the proposed framework, we evaluate communication performances given in terms of successful FC-based ARQ transmission probability and the average queueing delay of an underwater sensor. When evaluating the performances, we formulate the service time of each underwater sensor as a function of network parameters, i.e., the density of data sink and amount of redundancy for FC-based ARQ transmission, before solving a function for accurate service time, such that each sensor can be represented by an M/G/1 queue. Further, our analysis can formulate an optimization problem that aims at minimizing total cost incurred to install and operate 3D UASNs, without compromising two communication QoS requirements. To solve this problem, we propose a recursive algorithm to approach an optimal solution in reasonable time. Numerical evaluations demonstrate the validity of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In the cloud context, users are often called tenants. A cloud DBMS shared by many tenants is called a multi-tenant DBMS. The resource consolidation in such a DBMS allows the tenants to only pay for the resources that they consume, while providing the opportunity for the provider to increase its economic gain. For this, a Service Level Agreement (SLA) is usually established between the provider and a tenant. However, in the current systems, the SLA is often defined by the provider, while the tenant should agree with it before using the service. In addition, only the availability objective is described in the SLA, but not the performance objective. In this paper, an SLA negotiation framework is proposed, in which the provider and the tenant define the performance objective together in a fair way. To demonstrate the feasibility and the advantage of this framework, we evaluate its impact on query optimization. We formally define the problem by including the cost-efficiency aspect, we design a cost model and study the plan search space for this problem, we revise two search methods to adapt to the new context, and we propose a heuristic to solve the resource contention problem caused by concurrent queries of multiple tenants. We also conduct a performance evaluation to show that, our optimization approach (i.e., driven by the SLA) can be much more cost-effective than the traditional approach which always minimizes the query completion time. [ABSTRACT FROM AUTHOR]

Network functions virtualization (NFV) environments can provide increased elasticity and flexibility for operators, as network-related services can be scaled and moved as needed. Such operations require a seamless transfer of state to provide a service without interruption or performance degradation. In this paper, we propose and analyze a novel approach for efficient and seamless NFV state transfers. Our approach is based on the concept of statelets, which are compact representations of information in incoming packets that change the state of a virtualized network function. We present and describe slim migration (SliM), a system for efficient NFV state transfers using a statelet interface that we have implemented as an add-on to the data plane development kit, a high-performance packet I/O library. We have evaluated SliM in a testbed and present results that show its benefits in terms of lower delays and lower packet-loss rates. Moreover, our analysis and evaluation results show that SliM offers three times as much shared link capacity to the dataplane as previous approaches, while reducing the duration of a state transfer to one-third of their time. [ABSTRACT FROM PUBLISHER]

Due to high functional variety, wide power rating range, and relatively low product volumes, electronic systems for industrial application are difficult to economically be integrated on a chip. This paper presents the smart power gate array (SPGA), a novel application-specific integrated circuit (ASIC) platform exactly aimed at enabling cost-efficient single-chip integration of industrial electronic systems. The SPGA combines a low-voltage mixed-signal gate array with a structured high-voltage smart-power core, allowing for single-chip integration of signal conditioning, data processing, power management, actuator driving, and line communication functionalities required for industrial applications. Simultaneously, it offers low nonrecurring engineering cost and high design flexibility owning to the utilized gate array technology. In this paper, the SPGA concept is introduced to overcome integration limitations of existing ASIC technologies. Then, design and capabilities of the SPGA functional modules are described. Finally, a practical SPGA application example of single-chip integrated solenoid valve control and monitoring system is demonstrated. [ABSTRACT FROM PUBLISHER]

The fat-tree is one of the most widely-used topologies by interconnection network manufacturers. Recently, it has been demonstrated that a deterministic routing algorithm that optimally balances the network traffic can not only achieve almost the same performance than an adaptive routing algorithm but also outperforms it. On the other hand, fat-trees require a high number of switches with a non-negligible wiring complexity. In this paper, we propose replacing the fat-tree by a unidirectional multistage interconnection network (UMIN) that uses a traffic balancing deterministic routing algorithm. As a consequence, switch hardware is almost reduced to the half, decreasing, in this way, the power consumption, the arbitration complexity, the switch size itself, and the network cost. Preliminary evaluation results show that the UMIN with the load balancing scheme obtains lower latency than fat-tree for low and medium traffic loads. Furthermore, in networks with a high number of stages or with high radix switches, it obtains the same, or even higher, throughput than fat-tree. [ABSTRACT FROM PUBLISHER]