Client Architecture Research Articles

PerFedRLNAS: One-for-All Personalized Federated Neural Architecture Search

Personalized federated learning is a new paradigm to address heterogeneous problems (e.g. issues with non-i.i.d. data) in federated learning. However, existing personalized federated learning methods lack standards for how personalized and shared parts of the models are designed. Sometimes, manual design can even lead to worse performance than non-personalization. As a result, we propose a new algorithm for personalized federated neural architecture search, called PerFedRLNAS, to automatically personalize the architectures and weights of models on each client. With such an algorithm, we can solve the issues of low efficiency as well as failure to adapt to new search spaces in previous federated neural architecture search work. We further show that with automatically assigning different client architectures can solve heterogeneity of data distribution, efficiency and memory in federated learning. In our experiments, we empirically show that our framework shows much better performance with respect to personalized accuracy and overall time compared to state-of-the-art methods. Furthermore, PerFedRLNAS has a good generalization ability to new clients, and is easy to be deployed in practice.

Smartphone-Enabled Interaction on Large Displays—A Web-Technology-Based Approach

This paper proposes the design and implementation of a system for simultaneous communication of several users over a single display, such as large public displays or video walls. Motivated by the lack of software frameworks that are both free and suited for the desired level of complexity, we present a novel, out-of-the-box solution for collaborative and remotely controlled application use. The solution is based exclusively on web technologies, thus significantly facilitating usability and reach, and eliminating the need for additional software installations or the use of additional devices. The architecture of the developed solution is modular and comprises three main components: (1) the environment, (2) the remote controllable application (RCA), and (3) the controller. The environment serves as a central hub that manages the data flow and ensures synchronization between the user devices. The RCA, which runs on the large display, serves as the interface for interacting with the content, while the controller, which is loaded on the users’ smartphones, enables control and interaction with the RCA in real time. By using web sockets and a server–client architecture, the system supports device- and platform-independent remote control and centralized collaboration through an off-the-shelf communication framework entirely flexible regarding client application. In order to demonstrate the possibilities of the proposed system, several different scenarios were covered, and solutions were implemented and tested on a large display in the faculty lobby.

Hardware in the Loop Simulation for Bottle Sealing Process Virtualized on Unity 3D

This paper details the design and implementation of a virtualized bottle sealing plant using the Hardware in the Loop technique, for which it is divided into two parts: (i) Software consists of a virtualized environment in Unity 3D to visualize its behavior in real time; and (ii) Hardware was implemented through a PLC S7 1200 AC/DC/RLY (Programmable Logic Controller), which is responsible for the automation of the plant, programmed through the software TIA Portal V16 (Totally Integrated Automation Portal) and a control panel with buttons and indicator lights. The two developed parts communicate through bidirectional TCP/IP Ethernet, achieving a Server–Client architecture. For real-time monitoring and visualization, a SCADA (Supervisory Control and Data Acquisition) system implemented in InTouch is considered. In addition, the data acquisition is accomplished through the OPC (Open Platform Communication) server; the functionality of the OPC server is to transmit the information generated in an industrial plant at the enterprise level. This allows the process to execute its tasks of connectivity of automated processes and their supervision, as well as having scalability so that more tags can be included in other processes over time and ensure its operability.

Implementation of Novel Reduced-Order H∞ Filter for Simultaneous Detection and Mitigation of FDI-Attacks in AGC Systems

Recent developments in power systems to build a smart grid depend critically on communication networks which provide unprecedented advantages for comprehensive system visibility, monitoring, protection, and control. However, it increases the system complexity and vulnerability to cyberattacks that could jeopardize the seamless system operation. The attack must be detected effectively, and then proper measures to mitigate the effect quickly to restore normality. This work implements a reduced-order <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{\infty }$ </tex-math></inline-formula> filter to simultaneously estimate the state variable and unknown inputs such as noise or attack signals in the automatic generation control (AGC) system. Further, the knowledge of the disturbances is utilized to mitigate the attack by appropriately compensating the sensor readings, such that the control decisions are based on the correct states, not the corrupted signals generated in the attack scenario. The stability criterion is provided based on Lyapunov functions to check the estimation error in the presence of the exogenous disturbance. Further, the proposed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{\infty }$ </tex-math></inline-formula> estimator performance is evaluated by comparing it with the Kalman filter to verify its effectiveness. The estimator’s performance is validated by developing a cosimulation platform using the server–client architecture of data exchange. Further, the performance is validated in a real-time digital simulator (RTDS) platform used as the power system interacting with an estimator deployed on Raspberry Pi board sensing the attacks.

Web Mapping on Land Cover Change of Kathmandu, Lalitpur and Bhaktapur District

Land cover change is rapidly following path of environmental degradation, imbalance on ecosystem. Kathmandu, Lalitpur and Bhaktapur district has experienced a very high trend of urbanization and land use land cover change. Mapping this trend is absolutely necessary for sustainable development as well as conservation of the ecosystem. While there are desktop applications for making this mapping and analysis possible, upgrading to the web is the best possible way in order to interact with public as well as much easier way of data visualization, data analysis and data dissemination. This paper portrays the design and implementation process along with the visualization of a WebGIS platform via thin client architecture. A dynamic web map portal has been developed concerning the land cover change of Kathmandu, Lalitpur and Bhaktapur district. Among the several Free Open Source server software, OpenGeo Suite has been adopted for the project because it has a robust and flexible design that allows us to reliably handle and publish geospatial data. PostgreSQL/PostGIS, QGIS, Geoserver, Leaflet, and Apache Tomcat has been utilized as the web server in this suite. A fully functional website has been developed in order to host the maps and its components. WMS layers has also been published using Geoserver and styled as can be displayed on the web map portal. Leaflet plugins has been used with the help of which different tools like Zoom, Pan, Full Screen View and Distance Calculator are made available on the portal.

Public-Key Reinforced Blockchain Platform for Fog-IoT Network System Administration

The number of embedded devices that connect to a wireless network has been growing for the past decade. This interaction creates a network of Internet-of-Things (IoT) devices where data travel continuously. With the increase of devices and the need for the network to extend via fog computing, we have fog-based IoT networks. However, with more endpoints introduced to it, the network becomes open to malicious attackers. This work attempts to protect fog-based IoT networks by creating a platform that secures the endpoints through public-key encryption. The servers are allowed to mask the data packets shared within the network. To be able to track all of the encryption processes, we incorporated the use of permissioned blockchains. This technology completes the security layer by providing an immutable and automated data structure to function as a hyper ledger for the network. Each data transaction incorporates a handshake mechanism with the use of a public-key pair. This design guarantees that only devices that have proper access through the keys can use the network. Hence, management is made convenient and secure. The implementation of this platform is through a wireless server–client architecture to simulate the data transactions between devices. The conducted qualitative tests provide an in-depth feasibility investigation on the network’s levels of security. The results show the validity of the design as a means of fortifying the network against endpoint attacks.

A server bypass architecture for hopscotch hashing key–value store on DRAM-NVM memories

Non-volatile memories (NVMs), along with DRAMs, provide key–value (KV) stores with strong support in persisting data and storing it in memories. The remote direct memory access (RDMA) technology has been employed to boost remote data access to KV stores, since it provides kernel-bypass, zero-copy and low-latency features. Meanwhile, existing KV stores on DRAM-NVM memories suffer from performance bottlenecks; extra costs still need to be spent in accessing data in memories and keeping data consistency. This paper introduces Reno, a server bypass architecture for KV stores on DRAM-NVM memories. Reno is a server–client architecture for KV stores. It is equipped with two techniques: (1) a hopscotch hashing technique that allows KV stores on DRAM-NVM memories to perform latch-enabled append operations and (2) a fully server-bypass read/write paradigm. These techniques help eliminate network round trips of the server(s) and as well reduce hash conflicts, allowing clients to access the KV stores much more efficiently. We evaluate Reno on an Intel’s Optane DC Persistent Memory platform. The results show that Reno outperforms its counterparts by 1.1∼3.3× for remote reads and 1.9∼4.8× for remote writes in latency; the speedups of concurrent throughput are up to 2.33×, 2.36×, 3.09× and 5.08× for read-only, read-heavy, write-heavy and write-only YCSB workloads, respectively.

Manipulating vulnerability: Poisoning attacks and countermeasures in federated cloud–edge–client​ learning for image classification

The cloud–edge–client hierarchical computing architecture takes into account the dual advantages of cloud and edge computing. However, in realistic scenarios, the explosive increase of distributed data promotes the widespread application of the cloud–edge–client architecture, but also brings a variety of security threats. Relying on the advantages of federated learning technology, this has led to a growing interest in federated learning as the intersection of machine learning and cloud–edge–client fields. However, recent studies found that federated learning is potentially vulnerable to data poisoning attacks. Existing explorations of federated learning poisoning attacks and defense techniques in the cloud–edge–client environment are largely attack capabilities-specific and privacy protection-based: they are designed but fail to consider the characteristics of federated learning vulnerability and the difference in attack intensity, and lack flexibility. Yet few different attack capabilities and general defense strategies have been developed. To address these concerns, we introduce federated cloud–edge–client learning collaborative framework. Based on this architecture, a client-launched multi-label flipping attack method is discussed, which exhibits different attack capabilities. Further, we propose SPAD, a Server-based Parameter Adaptive Defense method in response to different attack intensities, to identify malicious attackers and benign participants, thereby improving the security of the federated learning image classification task. The results on image classification tasks demonstrate that in cloud–edge–client scenario, the multi-label flipping attack method presents different degrees of poisoning effect, and SPAD method has strong feasibility and effectiveness against poisoning attacks, which verify the significant advantages of the proposed method.

Edge-Assisted Democratized Learning Toward Federated Analytics

A recent take toward federated analytics (FA), which allows analytical insights of distributed data sets, reuses the federated learning (FL) infrastructure to evaluate the summary of model performances across the training devices. However, the current realization of FL adopts single server-multiple client architecture with limited scope for FA, which often results in learning models with poor generalization, i.e., an ability to handle new/unseen data, for real-world applications. Moreover, a hierarchical FL structure with distributed computing platforms demonstrates incoherent model performances at different aggregation levels. Therefore, we need to design a robust learning mechanism than the FL that 1) unleashes a viable infrastructure for FA and 2) trains learning models with better generalization capability. In this work, we adopt the novel democratized learning (Dem-AI) principles and designs to meet these objectives. First, we show the hierarchical learning structure of the proposed edge-assisted Dem-AI mechanism, namely <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Edge-DemLearn</i> , as a practical framework to empower generalization capability in support of FA. Second, we validate Edge-DemLearn as a flexible model training mechanism to build a distributed control and aggregation methodology in regions by leveraging the distributed computing infrastructure. The distributed edge computing servers construct regional models, minimize the communication loads, and ensure distributed data analytic application’s scalability. To that end, we adhere to a near-optimal two-sided many-to-one matching approach to handle the combinatorial constraints in Edge-DemLearn and solve it for fast knowledge acquisition with optimization of resource allocation and associations between multiple servers and devices. Extensive simulation results on real data sets demonstrate the effectiveness of the proposed methods.

WebMO: Web‐based computational chemistry calculations in education and research

AbstractWebMO is a web‐based interface for all major quantum chemistry programs. WebMO uses a server–client architecture that installs on a single server or cluster computer and provides access to state‐of‐the‐art computational chemistry programs from a standard web browser. The web interface provides a 3‐D molecular editor, pre‐defined calculations types, job submission and monitoring, visualization of results, and user management tools. Barriers to using state‐of‐the‐art computational chemistry in teaching and research are minimized through WebMO's universal accessibility, its intuitive and uniform interface to all programs, no software to install on client computers, and support for multiple users with a single instance. Applications of WebMO throughout the undergraduate curriculum are provided. The extensible open‐architecture design allows for collaboration among educators, researchers, quantum chemistry program developers, and the WebMO interface developers.This article is categorized under: Computer and Information Science &gt; Visualization Electronic Structure Theory &gt; Ab Initio Electronic Structure Methods Software &gt; Quantum Chemistry

Development of SOA-based WebGIS framework for education sector

The applications of Geographic Information System (GIS) in the education sector are increasing day by day. The geospatial information can be published, discovered, searched, analyzed, and displayed through webGIS-based applications. Lack of an open source geospatial resource-based platform for data sharing, discovery, and service delivery in the education sector is a critical issue in managing the education of large population in India. The use of open geospatial consortium (OGC) developed open standards for geospatial web services will result in the interoperability of geographic information. In this paper, an interoperable and secure service-oriented architecture (SOA)-based webGIS framework is developed to handle the technical and non-technical issues in the education sector. In this research work, spatial analysis on schools is performed along with the design and development of webGIS framework using SOA, OGC standards, and open source software. The developed webGIS framework, acronym as EduGIS, is interoperable and secure which is implemented for the education sector. The development of webGIS framework is based upon three-tier thin client architecture. The present research work has investigated an optimized adoption of various free and open source software like Quantum GIS, GeoServer, Apache Tomcat, PostGIS, and uDig in different tiers of developed webGIS framework. The interoperability of developed EduGIS ensures that it can be shared across different technologies, data, platforms, and organizations. The development of open source-based webGIS framework will serve as a means of reducing licensing costs in developing countries like India and will promote indigenous technological development for primary education in rural areas.

Development of Web Based Analysis Tool for Augmented Randomized Complete Block Design (ARCBD)

In biological and field experiments, the Augmented Randomized Complete Block Design (ARCBD) is widely used for screening and selection of a large number of germplasm lines/varieties/entries/test treatments with non replicated test treatments and replicated control treatments to estimate the experimental error. A web based online module for analysis of ARCBD was developed using scripting language Active Server Pages (ASP) based on server client architecture. The data have been taken from Federer (1956) and output compared accordingly. The outputs produced by the module are in agreement with the output generated from SAS package. An attempt was made to provide a user friendly interface for entering/pasting the data, characters names, number of observations and number of characters for analysis of augmented randomized complete block design. The module produces different output tables such as check x block table, block effects, control means and control effects, adjusted mean for test genotypes and genotypic effects. It also computes sum of squares in the analysis of variance tables after ignoring/eliminating treatment and eliminating/ignoring blocks for block and treatment effects, respectively. Critical difference table for comparing different mean differences at 5% and 1% level of significance is also given. A complete procedure is also provided in the help file to make a user friendly interface for analysis of the design.

SP-MIOV: A novel framework of shadow proxy based medical image online visualization in computing and storage resource restrained environments

The rapid development of Internet and various mobile communication media initiate the demands for access to medical image visualization systems. Medical image reading and interpretation at any time, any place and any device become an urgent need for radiologists. The current medical image online visualization methods have disadvantages in computing and storage resource restrained environments. This study presents a novel framework of medical image online visualization based on shadow proxy, which makes applications have across platform ability and universal environmental adaptability especial for devices with restricted running resources. The framework can be adapted in multiple client architectures including the pure web applications, mobile applications or regular desktop applications. It is easy to be integrated into third party software and there are no restrictions of the communication protocols between the client and server side due to two innovations of the framework that are shadow proxy mechanism and shadow data. The shadow proxy just does lightweight tasks on shadow data and the ultimate processing of computing tasks are moved to the server side to complete. The size of shadow data is small enough for shadow proxy that speeds up local display and processing tasks. Finally, the framework takes advantage of high performance on server side to render high quality image results. The performance of proposed work is evaluated in a web based medical image visualization system, and the results show that the framework in this paper allows the system to have smooth and quasi-real-time interaction performance. Therefore, this study ensures the local client operations fluency and fast while the quality of the visualization is still not lost that gives the best user experience.

Approaching the Communication Constraints of Ethereum-Based Decentralized Applications.

Those working on Blockchain technologies have described several new innovative directions and novel services in the Internet of things (IoT), including decentralized trust, trusted and verifiable execution of smart contracts, and machine-to-machine communications and automation that reach beyond the mere exchange of data. However, applying blockchain principles in the IoT is a challenge due to the constraints of the end devices. Because of fierce cost pressure, the hardware resources in these devices are usually reduced to the minimum necessary for operation. To achieve the high coverage needed, low bitrate mobile or wireless technologies are frequently applied, so the communication is often constrained, too. These constraints make the implementation of blockchain nodes for IoT as standalone end-devices impractical or even impossible. We therefore investigated possible design approaches to decentralized applications based on the Ethereum blockchain for the IoT. We proposed and evaluated three application architectures differing in communication, computation, storage, and security requirements. In a pilot setup we measured and analyzed the data traffic needed to run the blockchain clients and their applications. We found out that with the appropriate designs and the remote server architecture we can strongly reduce the storage and communication requirements imposed on devices, with predictable security implications. Periodic device traffic is reduced to 2400 B/s (HTTP) and 170 B/s (Websocket) from about 18 kB/s in the standalone-device full client architecture. A notification about a captured blockchain event and the corresponding verification resulted in about 2000 B of data. A transaction sent from the application to the client resulted in an about 500 B (HTTP) and 300 B message (Websocket). The key store location, which affects the serialization of a transaction, only had a small influence on the transaction-related data. Raw transaction messages were 45 B larger than when passing the JSON transaction objects. These findings provide directions for fog/cloud IoT application designers to avoid unrealistic expectations imposed upon their IoT devices and blockchain technologies, and enable them to select the appropriate system design according to the intended use case and system constraints. However, for very low bit-rate communication networks, new communication protocols for device to blockchain-client need to be considered.

Cloud-based improved Monte Carlo localization algorithm with robust orientation estimation for mobile robots

Purpose This paper aims to demonstrate a cloud-based version of the improved Monte Carlo localization algorithm with robust orientation estimation (IMCLROE). The purpose of this system is to increase the accuracy and efficiency of indoor robot localization. Design/methodology/approach The cloud-based IMCLROE is constructed with a cloud–client architecture that distributes computation between servers and a client robot. The system operates in two phases: in the offline phase, two maps are built under the MapReduce framework. This framework allows parallel and even distribution of map information to a cloud database in pre-described formats. In the online phase, an Apache HBase is adopted to calculate a pose in-memory and promptly send the result to the client robot. To demonstrate the efficiency of the cloud-based IMCLROE, a two-step experiment is conducted: first, a mobile robot implemented with a non-cloud IMCLROE and a UDOO single-board computer is tested for its efficiency on pose-estimation accuracy. Then, a cloud-based IMCLROE is implemented on a cloud–client architecture to demonstrate its efficiency on both pose-estimation accuracy and computation ability. Findings For indoor localization, the cloud-based IMCLROE is much more effective in acquiring pose-estimation accuracy and relieving computation burden than the non-cloud system. Originality/value The cloud-based IMCLROE achieves efficiency of indoor localization by using three innovative strategies: firstly, with the help of orientation estimation and weight calculation (OEWC), the system can sort out the best orientation. Secondly, the system reduces computation burden with map pre-caching. Thirdly, the cloud–client architecture distributes computation between the servers and client robot. Finally, the similar energy region (SER) technique provides a high-possibility region to the system, allowing the client robot to locate itself in a short time.

Visualizing Humanitarian Logistics Performance

Logistics management is crucial for the effectiveness and efficiency of humanitarian operations. Performance measurement enables practitioners to identify improvement potentials and management capabilities with regards to their logistics management tasks. While many performance measurement approaches for humanitarian logistics exist in the scientific domain, its applicability in the practitioner communities is rather low. Main reasons for this mismatch can be seen in rather few ready-to-use concepts and supporting information systems. In this article, the design, development, and evaluation of an information system for a Balanced Scorecard for humanitarian logistics is presented. The approach is embedded in the design science research framework following the agile programming methodology. The main characteristics of the server client architecture are described and reflected through experiences from formative and summative evaluations. The results stress the importance of the applied design approaches and support the closing of gaps between information systems designers and humanitarian practitioners.

A Magnetoencephalographic/Encephalographic (MEG/EEG) Brain-Computer Interface Driver for Interactive iOS Mobile Videogame Applications Utilizing the Hadoop Ecosystem, MongoDB, and Cassandra NoSQL Databases.

In Phase I, we collected data on five subjects yielding over 90% positive performance in Magnetoencephalographic (MEG) mid-and post-movement activity. In addition, a driver was developed that substituted the actions of the Brain Computer Interface (BCI) as mouse button presses for real-time use in visual simulations. The process was interfaced to a flight visualization demonstration utilizing left or right brainwave thought movement, the user experiences, the aircraft turning in the chosen direction, or on iOS Mobile Warfighter Videogame application. The BCI’s data analytics of a subject’s MEG brain waves and flight visualization performance videogame analytics were stored and analyzed using the Hadoop Ecosystem as a quick retrieval data warehouse. In Phase II portion of the project involves the Emotiv Encephalographic (EEG) Wireless Brain–Computer interfaces (BCIs) allow for people to establish a novel communication channel between the human brain and a machine, in this case, an iOS Mobile Application(s). The EEG BCI utilizes advanced and novel machine learning algorithms, as well as the Spark Directed Acyclic Graph (DAG), Cassandra NoSQL database environment, and also the competitor NoSQL MongoDB database for housing BCI analytics of subject’s response and users’ intent illustrated for both MEG/EEG brainwave signal acquisition. The wireless EEG signals that were acquired from the OpenVibe and the Emotiv EPOC headset can be connected via Bluetooth to an iPhone utilizing a thin Client architecture. The use of NoSQL databases were chosen because of its schema-less architecture and Map Reduce computational paradigm algorithm for housing a user’s brain signals from each referencing sensor. Thus, in the near future, if multiple users are playing on an online network connection and an MEG/EEG sensor fails, or if the connection is lost from the smartphone and the webserver due to low battery power or failed data transmission, it will not nullify the NoSQL document-oriented (MongoDB) or column-oriented Cassandra databases. Additionally, NoSQL databases have fast querying and indexing methodologies, which are perfect for online game analytics and technology. In Phase II, we collected data on five MEG subjects, yielding over 90% positive performance on iOS Mobile Applications with Objective-C and C++, however on EEG signals utilized on three subjects with the Emotiv wireless headsets and (n < 10) subjects from the OpenVibe EEG database the Variational Bayesian Factor Analysis Algorithm (VBFA) yielded below 60% performance and we are currently pursuing extending the VBFA algorithm to work in the time-frequency domain referred to as VBFA-TF to enhance EEG performance in the near future. The novel usage of NoSQL databases, Cassandra and MongoDB, were the primary main enhancements of the BCI Phase II MEG/EEG brain signal data acquisition, queries, and rapid analytics, with MapReduce and Spark DAG demonstrating future implications for next generation biometric MEG/EEG NoSQL databases.

Phaneros: Visibility‐based framework for massive peer‐to‐peer virtual environments

AbstractContemporary distributed virtual environments are growing out of terabytes of 3D content and hundreds of thousands of users. Server–client architectures have become inadequate for fulfilling the scalability requirements. The peer‐to‐peer architectures provide inherently scalable, cost‐effective distributed solutions for distributed virtual environments. We present a fully distributed peer‐to‐peer framework, Phaneros, which is capable of providing necessary means to realize more efficient and more scalable massive distributed virtual environments. Using the presented visibility‐aware interest management, Phaneros performs better than existing overlays, achieving single‐hop update dissemination while having lower bandwidth requirements. The provided visibility‐aware 3D streaming scheme distributes 3D content more efficiently without creating any significant load on the server. Our test results show significant improvements over existing frameworks.

A Taxonomy of IoT Client Architectures

This article defines a taxonomy of software architecture options, derived from industry projects, for Internet of Things (IoT) devices, from the most limited sensing devices to high-end devices featuring fully fledged OSs and developer frameworks. A plethora of architecture options exists for IoT devices, offering very different levels of software development capabilities. These capabilities can significantly affect IoT systems’ end-to-end architecture and topology.

English

Cloud computing is an internet based, remote driven and service oriented technology emerged to provide infrastructure as a service (IaaS), platform as a service (PaaS), software as a service (SaaS) among others. By implementing cloud computing, libraries will significantly reduce cost of maintaining systems, save energy, register a global outlook on the web, expose library patrons to multi-variant information resources and optimize their services. This study tries to provide understanding on the general relevance of cloud computing to libraries and how libraries can be transformed into smart institution in order to implement the cloud technology. Five facilities such as digital resources, internet service, thin client architecture, wireless access point and digital librarian are considered as prerequisites for cloud computing deployment. It is recommended that libraries can revive and maintain their relevance by integrating with the cloud computing technology, which will make them part of Internet of Things (IoT) experience that will soon dominate the activities of every educational sector. Key words: Cloud computing, library, information technology, thin client, wireless access point. &nbsp