Keynote 2: "Computing for big-data: Beyond CMOS and beyond Von-Neumann"

One of the most critical challenges for today's and future data-intensive and big-data problems is data storage and analysis. This paper first highlights some challenges of the new born Big Data paradigm and shows that the increase of the data size has already surpassed the capabilities of today's computation architectures suffering from the limited bandwidth, programmability overhead, energy inefficiency, and limited scalability. Thereafter, the paper introduces a new memristor-based architecture for data-intensive applications. The potential of such an architecture in solving data-intensive problems is illustrated by showing its capability to increase the computation efficiency, solving the communication bottleneck, reducing the leakage currents, etc. Finally, the paper discusses why memristor technology is very suitable for the realization of such an architecture; using memristors to implement dual functions (storage and logic) is illustrated.

In the recent decades, remote sensing data are rapidly growing in size and variety, and considered as “big geo data” because of their huge data volume, significant heterogeneity and challenge of fast analysis. In the traditional remote sensing analysis workflows, the data transfer for downloading raw image files to local workstations often costs a lot of time and slows down the data analysis workflows. Because results of remote sensing data analysis models are usually much smaller than raw data to be processed, “on-demand processing”, which tries to upload data analysis models and execute them “near” where data stores, can significantly accelerate the execution of remote sensing analysis workflows. In this paper, a framework for on-demand remote sensing data analysis is proposed based on three-layered architecture; XML/JSON based runtime environment description; and on-demand model deployment methods. The evaluation on a prototype system shows that on-demand processing framework accelerates the execution of analysis models in 2.8 ~ 12.7 times by reducing data transfers, especially for those analysis workflows which transfer data through low bandwidth Internet. By on-demand processing, classical remote sensing data service systems can evolve into remote sensing data processing infrastructures, which provide IaaS (Infrastructure-as-a-Service) and PaaS (Platform-as-a-Service) services, and make it possible to exchange knowledge among scientists by sharing models. Furthermore, a remote sensing data analysis platform for carbon satellites is designed based on the on-demand processing proposed by this paper and will soon be implemented under the support of SunWay-TaihuLight, the world’s most powerful super computer.

The increasing availability of Earth Observation and geographic data implies great opportunities for those capable to efficiently address the problems arising from the management of this huge amount of data. An efficient management of these data means to respond to the paradigm of the 4 V that usually applies to the problem of the Big Data: Volume - the sheer size of the “data at rest”, Velocity - the speed of new data arriving, Variety - the different manifold, and Veracity - trustworthiness and issues of provenance. The problem of Big Data is not only in the geospatial realm of Earth observation data, but in general in all the location-based data which basically are the main contributors to the deluge of big data. In addition to the need of managing (store, search & find when needed) these data efficiently, the problem arises from the analysis of these data. They need to be quickly processed in order to quickly extract the information content, then they must be analysed in conjunction with other data sources in order to express their real value in the construction of new knowledge. These processes are hastened by the advent of an increasing machine-to-machine communication. The automation of the data analysis requires standardized and linked data so that they can be processed by machines without human intervention. The problem with the standardization of geospatial data is solved by simply observing not only the best practices shared at European level, but mainly the regulatory scenario dictated by the INSPIRE Directive . The publication of spatial data as Linked Open Data may then leverage the reuse of common ontologies and vocabularies that allow the connection of geospatial data with other heterogeneous information. This way new scenarios and business opportunities may arise, as in the case of the real estate market that is mentioned in this article. This contribution aims to identify some business opportunities, related to Linked Open Data and arising from the imminent availability of the Sentinel satellite data, with the European program Copernicus, for companies operating in the so-called downstream services of Earth observation.

Big Data is a very huge volume of data which is beyond the storage capacity and processing capability of traditional system. The volume of data is increasing at exponential rate. Therefore, there is the need of such mechanism to store and process such high volume of data. The impressiveness of the Big data lies with its major applicability to almost all industries. Therefore, it represents both, the tremendous opportunities and complex challenges. Such omnipotent eminence leads to the privacy and security related challenges to the big data. Nowadays, security of big data is mainly focused by every organization because it contains a lot of sensitive data and useful information for taking decisions. The hostile nature of digital data itself has certain inherited security challenges. The aim of Big data security is to identify security issues and to find the better solution for handling security challenges. The observation and analysis of different security mechanism related to the issues of big data and their solutions are focused in this chapter.

Creating intelligence information products is a time consuming and difficult process for analysts faced with identifying key pieces of information relevant to a complex set of information requirements. Complicating matters, these key pieces of information exist in multiple modalities scattered across data stores, buried in huge volumes of data. This results in the current predicament analysts find themselves; information retrieval and management consumes huge amounts of time that could be better spent performing analysis. The persistent growth in data accumulation rates will only increase the amount of time spent on these tasks without a significant advance in automated solutions for information product generation. We present a product generation tool, Automated PrOduct Generation and Enrichment (APOGEE), which aims to automate the information product creation process in order to shift the bulk of the analysts’ effort from data discovery and management to analysis. APOGEE discovers relevant text, imagery, video, and audio for inclusion in information products using semantic and statistical models of unstructured content. APOGEEs mixed-initiative interface, supported by highly responsive backend mechanisms, allows analysts to dynamically control the product generation process ensuring a maximally relevant result. The combination of these capabilities results in significant reductions in the time it takes analysts to produce information products while helping to increase the overall coverage. Through evaluation with a domain expert, APOGEE has been shown the potential to cut down the time for product generation by 20x. The result is a flexible end-to-end system that can be rapidly deployed in new operational settings.

As the number of patients treated in-home are increasing exponentially mainly in countries such as Japan, US A and Europe etc., these people often enter into a critical situation that may require help immediately (e.g. when facing an accident, or becoming depressed). Researches and Advances in computing areas and the Internet of Things (IOT) have provided efficient and cheap equipments including wireless communication and cameras, such as smart phones or embedded devices that enable the deployment of Health S mart Homes (HS H) that can provide medical treatment of patients in their homes. The images captured from these cameras can help nurses or caretakers of patients to provide timely help. The use of such patient images and emotional detection to assist patients and elderly people within home is provided in this article. Internet of Things and Cloud computing can work together, which can address to the Big Data problems. Big data is actually a term used for a huge data set that performs operations like storage of data, analysis, sharing, transfer, predictive analysis; updating etc. data set can grow rapidly. Data sets have different analytics on data that involve process of inspecting, transformation and modeling of data so as to discover new information, and to reach on some particular decision. There are some particular data analysis techniques from which data mining is a popular one which focuses on modeling and discovery of new facts. The discovered knowledge helps in predictive analysis purposes as well as text analysis. Further on data mining is performed on data set which involves discovery of patterns by applying different operations at databases, and using artificial intelligence, machine learning, statics etc. Keywords: Cloud computing, IOT, AI, face recognition, big data, data mining.

Near-data processing is a promising paradigm to address the bandwidth, latency, and energy limitations in today's computer systems. In this work, we introduce an architecture that enhances a contemporary multi-core CPU with new features for supporting a seamless integration of near-data processing capabilities. Crucial aspects such as coherency, data placement, communication, address translation, and the programming model are discussed. The essential components, as well as a system simulator, are realized in hardware and software. Results for the important Graph500 benchmark show a 1.5x speedup when using the proposed architecture.

Transformer model has achieved excellent results in many fields, owing of its huge data volume and high precision requirements, the traditional analog compute-in-memory circuit can no longer meet its needs. To solve this dilemma, this paper proposes a digital compute-in-memory circuit based on the improved Booth algorithm. The 6T SRAM array stores the multiplicand, and the multiplier is encoded by the booth encoder, and then, local computing cell (LCC) read the corresponding value from the array according to the encoding result. These values are finally sent to the dual-mode shift and add module (DMSA) to obtain the computation results. The proposed circuit achieved energy efficiency of 33.11TOPS/W@INT8 and 8.3 TOPS/W@INT16. And the proposed circuit achieved 1.92+ better energy efficiency compared with previous works.

In recent years, the researchers have got progressively more interest in wireless sensor networks (WSNs) as it plays a vital role in numerous applications. WSNs have strict resource constraints such as limited bandwidth, limited energy, limited processing capability, and limited memory. Among them, the most critical one is limited energy. So, energy efficiency has got significant importance in the WSNs. Energy consumption is mainly due to computation and communication between sensor nodes and the base station. Energy consumption for communication is more when compared to computation. Hence it is indispensable to focus on designing an energy-efficient routing for WSN. In this paper, a cluster-based routing using simulated annealing and genetic algorithm-based hybrid (SAGA-H) approach has been proposed. The proposed approach is explained and simulated using MATLAB. Further, the results observed have been compared with existing GA-based approach with respect to network lifetime, the number of packets sent to BS and sink, and average residual energy. From the simulation results, it is observed that the proposed approach outperforms the existing GA-based approach.

Crystal chips mimic brain to sift giant piles of data

Nowadays a huge volume of biomedical data (images, genes, etc) are daily generated. The interpretation of such data involves a considerable expertise. The misinterpretation and/or misdetection of a suspicious clinical finding leads to increasing the negligence claims, and redundant procedures (e.g. biopsies). The analysis of biomedical data is a complex task which are performed by specialists on whose expertise degree affects the accuracy of their diagnosis. Besides, due to the huge volume of data, it is a tiresome process. To mitigate these intrinsic drawbacks Computeraided Diagnosis approaches have been proposed in the last decade, but applied without a deep analysis. It is also very common in the literature for the presentation of experimental results to rely solely on the mean of accuracy values. This procedure is not always reliable, especially for applications that require faster classifiers due to their learning-time constraints. Hence, in this paper we proposed an extensive analysis towards an effective and efficient learning for biomedical data classification. To do so, several public biomedical datasets were used against different supervised classifiers, taking into account accuracies and computational times obtained throughout the learning process.

As today's computer architectures are becoming more and more heterogeneous, a plethora of options including CPUs, GPUs, DSPs, reconfigurable logic (FPGAs), and other application-specific processors come into consideration for close-to-sensor processing. Especially, in the domain of image processing on mobile devices, among numerous design challenges, a very stringent energy budget is of utmost importance, making embedded GPUs and FPGAs ideal targets for implementation. Recently, the HIPAcc framework was proposed as a means for automatic code generation of image processing algorithms for embedded GPUs, based on a Domain-Specific Language (DSL). Despite of huge advancements in High-Level Synthesis (HLS) for FPGAs, designers are still required to have detailed knowledge about coding techniques and the targeted architecture to achieve efficient solutions. As a remedy, in this work, we propose code generation techniques for C-based HLS from a common high-level DSL description targeting FPGAs. Our approach includes FPGA-specific memory architectures for handling point and local operators, numerous high-level transformations, and automatic test bench generation. We evaluate our approach by comparing the resulting hardware accelerators to existing frameworks in terms of performance and resource requirements. Moreover, we assess the achieved energy efficiency in contrast to software implementations, generated by HIPAcc from the same code base, executed on GPUs.

Charting the path forward: CT image quality assessment - an in-depth review

Due to its historical nature, data warehouses require that large volumes of data need to be stored in their repositories. Some organizations are beginning to have problems to manage and analyze these huge volumes of data. This is due, in large part, to the relational databases which are the primary method of data storage in a data warehouse, and start underperforming, crumbling under the weight of the data stored. In opposition to these systems, arise the NoSQL databases that are associated with the storage of very large volumes of data inherent to the Big Data paradigm. Thus, this article focuses on the study of the feasibility and the implications of the adoption of a NoSQL database, within the data warehousing context. MongoDB was selected to represent the NoSQL systems in this investigation. In this paper will be explained the processes required to design the structure of a data warehouse and typically dimensional queries in the MongoDB system. The undertaken research culminates in the performance analysis of queries executed in a traditional data warehouse, based on the SQL Server system, and an equivalent data warehouse based on the MongoDB system.

Continuous improvement in hardware design and advances in wireless communication have enabled the deployment of various wireless sensor applications. Though, sensor nodes are highly resource constrained in terms of limited processing speed, runtime memory, persistent storage, communication bandwidth and finite energy. Flash memory has become a more widespread storage medium for modern wireless devices because of its effective features like nonvolatility, small size, light weight, fast access speed, shock resistance, high reliability and low power consumption. Even though flash memory offers many advantages, its special hardware read, write and erase characteristics impose design challenges on storage systems. This survey analysis the flash based storage management techniques proposed for resource constraint sensor devices, in terms of data life, device life, energy, and storage efficiency. Comprehensive simulation results show gradually improved performance in area of wireless sensor networks.