International Defense Security & Technology Your trusted Source for News, Research and Analysis
Related Articles
Over the past decades, cloud computing has been greatly developed and applied owing to its high cost-efficiency and flexibility achieved through consolidation, in which computing, storage, and network management functions work in a centralized manner. The increase of IoT devices at the edge of the network is producing a massive amount of data to be computed at data centers, pushing network bandwidth requirements to the limit. Despite the improvements of network technology, data centers cannot guarantee acceptable transfer rates and response times, which could be a critical requirement for many applications. Mobile devices connected to distant centralized cloud servers try to obtain sophisticated applications, which impose additional load on both Radio Access Networks (RANs) and backhaul networks and result in high latency.
It’s estimated that by 2025, 75 percent of data will be created outside of central data centers, where most processing takes place today. Taking this a step further, approximately 90 percent of all data collected by enterprises today will never be used. Edge computing provides a path to reap the benefits of data collected from devices through high-performance processing, low-latency connectivity, and secure platforms.
The emerging IoT introduces new challenges, such as stringent latency, capacity constraints, resource-constrained devices, uninterrupted services with intermittent connectivity, and enhanced security, which cannot be adequately addressed by the centralized cloud computing architecture.
Next-generation networks will require the support of interactive AI-powered services and some services like autonomous vehicles are sensitive to response latency, which needs to interact intelligently with their environments in real-time. A promising solution is known as “edge computing” is emerging that refers to the storage, processing, and analysis of data nearer to the edge of a user’s network, wherein the data is generated to enable rapid, near real-time analysis and response. Edge computing technology moves the computation away from centralized data centers by exploiting smart objects, mobile phones, or network gateways to perform tasks and provide services on behalf of the cloud.
Edge computing is a networking solution that reduces the number of processes running on the cloud and moves them to local devices, such as the user’s computer, an IoT device, or an edge server. It minimizes the amount of long-distance communication between the client and the server, which decreases latency and improves process efficiency. Consequently, several organizations are adopting edge computing as it lowers bandwidth use, associated costs, and server resources
Such a capability would allow us to analyze this information effectively and, in turn, discover solutions to some of our most pressing problems, from traffic congestion to the spread of disease to clean energy alternatives. This offers superior control and management of the data, while cutting down on operational costs.
Securing sensitive data, such as private medical records, at the edge and transmitting less data across the internet could help increase security by reducing the risk of interception. In addition, some governments or customers may require that data remain in the jurisdiction where it was created. In healthcare, for example, there may even be local or regional requirements to limit the storage or transmission of personal data.
Lack of persistent internet connectivity can impede cloud computing, but a variety of network connectivity options make edge-to-cloud computing feasible. For example, 5G provides a high-bandwidth, low-latency connection for rapid data transfer and service delivery from the edge. The amount of data that networks can transmit at one time is limited. For locations with subpar internet connectivity, being able to store and process data at the edge improves reliability when the cloud connection is disrupted.
Edge computing is more suitable to be integrated with IoT to provide efficient and secure services for a large number of end-users, and edge computing-based architecture can be considered for the future IoT infrastructure. IoT and edge computing devices collect data and manage it in one of two main ways. Intelligent edge computing devices with built-in processors may offer advanced capabilities like analytics or AI onboard, while devices without processors send the data they generate to a server deployed at the on-premises edge for storage and analysis.
An on-premises edge server can then process data from the edge computing devices and return critical information needed for near real-time applications or send only the relevant portions of the data to the cloud. Data from numerous edge computing devices can be consolidated in the cloud for more extensive processing and analysis.
Edge Computing technology
Edge computing utilizes the distributed technology, including distributed networking, computing nodes, storage resources as well as hardware safety control units and technological integration present in devices and equipment. Devices or equipment for edge computing prioritize the processing of data or information from multiple end nodes or equipment and then send the data or information back to the cloud environment. Contrary to the traditional basic cloud computing environment, it uses the concept of the centralized computing environment in which all data are analyzed and processed on the cloud.
The key technologies of Mobile edge computing ( MEC) including computation offloading and mobility management. Computation offloading is a procedure that migrates resource-intensive computations from a mobile device to the resource-rich nearby infrastructure. Although mobile devices are constrained by computing capabilities, battery life, and heat dissipation, MEC enables running new sophisticated applications at UEs by offloading energy-consuming computations of the applications to the MEC server. MEC applications are running as VM on top of the virtualization infrastructure, and can interact with the mobile edge platform to perform certain support procedure.
To address this challenge, the cloudlet, which is a mobility-enhanced small-scale cloud Data Center (DC) that is located at the edge of the internet, is proposed. A cloudlet is a trusted, resource-rich computer or cluster of computers that are well-connected to the internet and available for use by nearby mobile devices . The main purpose of the cloudlet is supporting resource-intensive and interactive mobile applications by providing powerful computing resources to mobile devices with lower latency. User Equipments (UEs) can access the computing resources in the nearby cloudlet through a one-hop high-speed wireless local area network. Cloudlets represent the middle tier of the 3-tier hierarchy architecture (mobile device layer, cloudlet layer, and cloud layer) to achieve crisp response time. By incorporating a local high-performance processor with built-in artificial intelligence (AI), local decision-making can be carried out and only if necessary, communicated with the cloud.
EC nodes are envisioned to perform tasks, such as real-time signal and image processing, combinatorial optimization, agent-based modeling, big data analysis, etc. Such tasks are performed to provide secured services, effective control, and seamless decision-making while achieving energy efficiency. Therefore, HPC is essential to EC networks. Thus, physically, servers are deployed at the near edge or the extreme edge of the network (closer proximity to the data sources) instead of the data centers. By incorporating a local high-performance processor with built-in artificial intelligence (AI), local decision-making can be carried out and only if necessary, communicated with the cloud.
Although currently, supercomputers, such as Summit and Sierra (ranking No. 1 and No. 2, respectively on the Top500 list) are not practical for EC, it is envisioned that smart supercomputers will eventually enter the EC domain, unleashing new capabilities. Summit, for example, a US Dep. of Energy machine at Oak Ridge National Laboratory, links more than 27,000 GPUs and 9000 CPUs to provide 200 petaflops (=2 × 1017 operations/s) HPC at a power consumption of 13 MW, which is not practical yet for EC devices.
Within the IoT, including IIoT (industrial IoT ) and IoMT (internet of medical things ), the EC devices are extremely diverse, and the volume of data they are generating and processing is rapidly increasing. Data formats include time and frequency space signals, complex images, sound and voice, and a plethora of protected health, personal, and sensitive data. Due to the variety of EC devices, data types, and algorithms, many AI-based or smart offloading and transmission strategies are being proposed, such as employing machine and deep learning methods, or mimicking human brain networks
Increasingly more complex services are expected to be provided by the continuously improving energy efficiency of system-on-chip (SoC), furnishing sensors and devices (edge, mist, fog, and IoT end devices) with significant computing power. This allows a self-contained module in which edge sensors and actuators generate data that can be processed on-site. In addition to software, SoCs, containing components, such as processors, graphics processing unit GPU), network-on-chip (NoC), and memory and data storage, essentially function as a server. Attractive/desired attributes of SoCs include their SWaP (size, weight, and power).
The power efficiency of EC devices is thus of paramount importance despite the development of better energy storage and power transport technologies. Designers, developers, and manufacturers compete to achieve smaller, lighter, lower cost, but faster, higher performance, and more energy-efficient processors for EC applications. Thus, as the EC-use cases are being more systematically characterized, the design of EC-optimized processors is expected to intensify.
Processor design begins by considering a specific instruction set architecture (ISA), which provides the needed information to write machine language programs and implement different processors. The ISAs, having varying degrees of complexity, can lead to processors that may be more suitable for EC and edge-native applications. For example, an ISA that retains specialized instructions, including those that may be used less frequently in practical programs, is the complex instruction set computer (CISC), which was the basis for Intel’s 80 × 86 chip. CISC, however, has not been seen as competitive for EC, which focuses more on specific performance and functionality criteria.
Since RISC processors carry out fewer computer instructions, they operate faster, which is of specific importance in many EC devices where the real-time or simultaneous response is desired. By excluding instructions that are not needed, RISC processors, employing a reduced number of transistors, use a fraction of the power required by CISC processors. RISC processors can also be more suitable for miniaturized EC devices since the size of the semiconducting material (die size) needed for the integrated circuit is proportional to the transistor count, which leads to smaller processors.
To support user mobility, a VM handoff technology needs to be used to seamlessly migrate the offloaded services on the first cloudlet to the second cloudlet as a user moves away from the currently associated cloudlet. When the UE performs a handover to another cell, it is important to guarantee the service continuity and QoS requirements . VM migration is a service included in many hyper-visors to move VMs from one physical machine to another, commonly within a DC.
As discussed by D. Patterson in “Reduced Instruction Set Computers Then and Now” , RISC grew out of an attempt to execute more instructions in a single short cycle. With the rapidly advancing IoT applications, placing increasing demands on EC, new and innovative ISAs, such as the open-source RISC five (RISC-V) , is paving the way for new architectures, allowing hardware designers to implement powerful processor for both EC and the cloud. With frozen base instructions (while supporting custom instructions for designing specialty functions), software written for RISC-V will indefinitely run on other similar RISC-V cores.
Examples of new commercial processors for EC applications include chips from the Advanced RISC Machine (ARM). In addition to products such as the Cortex families, the recently announced Neoverse solutions are explicitly advertised for EC-use cases. The main characteristics that advertise these new chips within the EC domain include low latency, low power consumption, and smaller size. Others offering or competing to supply EC hardware, capabilities, and services include NVIDIA EGX platform, APC’s Edge Computing Solutions, Open Edge Computing Initiative, and others.
IBM, mimik technology Partner to Advance Edge Computing
Canada-based mimik technology specializes in hybrid edge cloud computing. The company’s hybrid edgeCloud platform enables to transform any computing device to work as a cloud server. This helps application developers accelerate app development. The start-up is headquartered in Vancouver and founded in 2009.
mimik technology is a participant of IBM’s edge ecosystem. The tech giant’s edge ecosystem assists in the deployment of open standards-based cloud-native solutions for equipment manufacturers, IT & software providers. These cloud-native solutions are capable of autonomously managing edge applications at scale. mimik technology noted that due to the rapid proliferation of technologies like IoT and 5G as well as rapid migration to cloud has resulted in increasing number of smart devices at the edge. This is exerting enormous pressure on computing resources and internet bandwidth of the centralized data centers. Also, the increasing complex nature of clients’ application is making it difficult to update and manage such applications.
IBM is also expanding its presence in the edge computing market to gain a larger share. Its IBM Edge Application Manager platform leverages the Red Hat OpenShift platform and aids in autonomous management for edge computing. The platform was unveiled in May 2020. Per company estimates, the platform is capable of allowing a single administrator to manage up to 10,000 edge nodes at a time. Further, IBM’s portfolio of other edge-enabled applications and services include IBM Asset Optimization, IBM Visual Insights, IBM Production Optimization, IBM Connected Manufacturing, IBM Maximo Worker Insights and IBM Visual Inspector.
The combination of mimik technology’s hybrid edgeCloud platform and IBM’s Edge Application Manager will provide containerized computing functionalities at the edge for a wide range of Linux-powered edge devices and clusters. The integration will also offer containerized computing capabilities at the edge for various devices like smartphones, PCs, routers, and IoT gateways, running on different operating systems like Android, iOS, Windows and QNX.
Nanosystems and Nanoscience: From Edge Sensing to Edge Computing
ORNL researchers Ali Passian and Neena Imam have surveyed the edge computing landscape, as well as novel nanoscale technologies, to better understand how to simultaneously advance both edge computing and nanoscience to benefit scientific progress. Their work was published in the journal Sensors.
Computing has successfully capitalized on the electronic properties of silicon and silicon-based electronics. The success builds primarily upon the scalability of silicon transistors, such as a field-effect transistor (FET), which is the basic component of present computer circuitry. To surpass sub-20 nm nodes, that is, the semiconductor manufacturing process that generates transistors with a size smaller than 20 nm, the conventional scaling has reached major limitations. Currently, silicon FETs are fabricated at the 14-nm node (and at the sub-14-nm node using FinFET technology) and have an overall lateral footprint of ~90–100 nm. Innovative approaches and extreme ultraviolet lithography to print the features are being explored towards 5 nm node design. Thus, nanoscale phenomena are expected to be manifested more strongly in the increasingly nanosized domains. In nanosystems, because the electronic and optical response of bulk materials are altered by size, shape, and surface, confinement effects can be pronounced and thus provide new functionalities.
In the general case of a composite nanostructured material domain, such as a quantum well of a given morphology or a gap-plasmonic structure, the ensuing quantum confinement furnishes a variety of enabling mechanisms via tunneling, modification of local density of states, frustrated total internal reflection, mode coupling, etc. It is envisioned that the opportunities offered by fields, such as advanced nanophotonics, nanomaterials, and smart sensors, will be capitalized by EC to advance IoT and other network-based applications. Nanosystems, such as nanophotonic crystal cavities, quantum dots, carbon nanotubes (CNT), and nanomaterials, and their composites allow information processing. Owing to the unique properties of nanomaterials and nanostructures, future processors and integrated circuits will exploit these nanosystems not only for computing and conveying information but also for storage of information.
From the perspective of novel systems of potential for EC-specialized applications, discerning the technological challenges in achieving faster, smaller, and more energy-efficient processors and components, including interconnects, storage, communication, and software, helps to define the broader scope of the EC as a field. Clearly, despite the growing number of reports that are sharpening the boundaries of the EC field, the cross-disciplinary nature of the field must also be considered. This is not unlike the cross-disciplinary field of nanoscience, which provides the potential hardware and implementation solutions for EC, as described in the case of CNT-based transistors for post-Moore needs. However, in turn, emerging unique EC-use cases also provide new challenges for nanoscience. For example, from the presented discussion on CNT synaptic transistors for neuromorphic computing, it may be surmised that nanosystems and EC may amalgamize to become an inseparable entity, where device and function interact dynamically.
Measuring the global needs for better and new sensors, it can readily be concluded that increasing the computing power and the intelligence of the sensors is only a natural evolution of the related industries. The majority of EC devices are expected to be fast, compact, low power, and resilient to becoming compromised (hardware and software). The processor unit that could give the EC sensors and devices the needed intelligence to provide an “edge” advantage needs also meet these same criteria.
Remarkably, EC appears to impact a plethora of scientific and technological fields as it overcomes many challenges, such as the potential risk for increased hacking vector and licensing costs. Clearly, important challenges remain to be addressed with respect to security, software upgrade, and others related to the touted EC advantages.
The utility of light and optical excitations in various nanosystems is of tremendous importance for EC. In general, communication may be regarded as a transfer or exchange of information. Modulation is the essence of communication. Measurable variations in a quantity, such as the electric field amplitude, phase, or polarization, can provide a route to modulation and thus can convey information.
Development of all-optical components, photonic chips, interconnects, and processors will bring the speed of light, photon coherence properties, field confinement and enhancement, information-carrying capacity, and the broad spectrum of light into the high-performance computing, the internet of things, and industries related to cloud, fog, and recently edge computing. Conversely, owing to their extraordinary properties, 0D, 1D, and 2D materials are being explored as a physical basis for the next generation of logic components and processors. Carbon nanotubes, for example, have been recently used to create a new processor beyond proof of principle. These developments, in conjunction with neuromorphic and quantum computing, are envisioned to maintain the growth of computing power beyond the projected plateau for silicon technology.
Other uses of polymeric nanomaterials include shape memory functionality, explored by IBM (international business machines) corporation through the “millipede” system aimed at developing ultrahigh density storage devices with terabit capacity, small form factor, and high data rate
The role of nanomaterials and nanosystems in overcoming these obstacles takes the center stage. Not only the hardware that accommodates the computing power, intelligence, and communication capacity of the EC devices but also the signal generating sensors and controlling actuators benefit from the extraordinary properties of the reviewed materials. With the rising fusion of the IoT devices globally, EC could help to unburden the resulting colossal computing and communication loads. In turn, the evolution of EC devices and the synergy among the sensing, computing, and AI components are expected to enable new scientific and technological capabilities.
As the EC device dimensions are pushed to smaller scales to achieve higher density integration, the boundaries of nanosystems and EC further merge. It may be argued that the state of the nanosystems-EC field is only at its infancy with such perspectives as collective and holistic responses from high connectivity, synchronized, and real-time EC devices may be tamed to provide new horizons of information technological capabilities. Mathematical problems, for the spatial distribution of edge devices, or the rate of computing and data exchange, etc. may be formulated, to better understand the underlying complexity, and thus emergence.
EC will reach its climax as novel nanosystems that switch fast, dissipate less energy per switching, implement functions, transport information-carrying signals fast and with little dissipation per unit length are discovered and harnessed. Currently, materials that exhibit topological and quantum behavior, metamaterials, and nanoparticles and nanowires that can be integrated are under consideration.
The prime example of 1D and 2D nanomaterials of current interest within the enormous application space, comprising next-generation computing systems and circuit elements, includes CNTs and graphene, MoS2, transition metal dichalcogenides, black phosphorus, etc. Building processors based on carbon nanomaterial FETs (MOSFET, CMOS, and the multi-gate transistors FinFET have been already demonstrated as in the case of CNT FET. Advanced FETs, such as CNT-based FinFETs are being explored as a power-efficient node-scaled platform towards a chip with reduced transistor dimensions and thus increased density. Use of other nanomaterials of relevance includes fabrication and testing via self-assembly of block copolymers to achieve 7 nm node FinFETs, and Si, Ge, and SiGe nanowire FinFETs
Stimulated by successful demonstrations, such as trapping a single atom or imaging single atomic sites and bonds within a molecule, there have been visions to reach beyond nanotechnology for continuing at thousand times smaller into the pico-technology, and at a million times smaller into the femto-technology, the realm of neutrons, protons, electrons, and other nuclear particles. Amazingly, pico- and femto-technologies are already being contemplated for addressing technology bottlenecks, such as a better electronic on-off switching speed to improve communication bandwidths beyond ~50 GHz
Recently the notion of molecules themselves being used as computers, or quantum effects being employed to compute and communicate, have emerged. Astonishingly, DNA computing, molecular machines, biological microprocessors, bio-electronic computers, etc. have been already reported, albeit largely exploratory.
With evolving nanosystems, one may envision molecular systems, e.g., proteins and DNAs, as EC devices. With sensing at the atomic and molecular levels, nanoscale communication, molecular processing, and nano-EC devices may pave the way to nano-IoT. Molecular networks of billions of sensors already occur in biological systems, and this may be mimicked by nano-EC devices.
Taming individual atoms towards quantum computing, atom-by-atom assemblers to arrange several trapped neutral atoms in one-dimension , in arbitrary two-dimensional patterns, and in three-dimensional arrays with controllable single atom capability, have been demonstrated. To scale up the “fabrication” of such atomic and molecular switches, novel concepts are being reported, including the demonstration of monolayer surface patterning at 3.5 nm on a gold surface via self-assembly, offering a potential path to large-area patterning .
