Homogeneous Hardware Research Articles

A dual-mode organic memristor for coordinated visual perceptive computing

The hierarchically coordinated processing of visual information with the data degradation characteristic embodies the energy consumption minimization and signal transmission efficiency maximization of brain activities. This inspires machine vision to handle the explosively increased data in real-time. In this contribution, we demonstrate the possibility of constructing a coordinated perceptive computing paradigm with dual-mode organic memristors to emulate the visual processing capability of the brain systems. The 32-state modulation of the device photoresponsivity and conductance via photo-induced molecular reconfiguration and electrochemical redox activities enables the execution of computing-in-sensor and computing-in-memory tasks, respectively, which in turn allows the homogeneous hardware integration of a single-layer perceptron and a convolutional neural network for high-efficiency hierarchical visual processing. Compared to the sole optoelectronic CIS mode to recognize visual targets, the dual-mode organic memristor-based coordinated computing scheme demonstrates a 24.5% improvement in the recognition accuracy and 45.8% reduction in the network size.

Control of large helium cryogenic systems: a case study on CERN LHC

The CERN cryogenic facilities demand a versatile, distributed, homogeneous and highly reliable control system. For this purpose, CERN conceived and developed several frameworks (JCOP, UNICOS, FESA, CMW), based on current industrial technologies and COTS equipment, such as PC, PLC and SCADA systems complying with the requested constraints. The cryogenic control system nowadays uses these frameworks and allows the joint development of supervision and control layers by defining a common structure for specifications and code documentation. Another important advantage of the CERN frameworks is the possibility to integrate different control systems into a large technical system with communication capability. Such a system is capable of sharing control variables from all accelerator apparatus in order to cope with the operation scenarios.The first implementation of this control architecture started in 2000 for the Large Hadron Collider (LHC). Since then CERN continued developing the hardware and software components of the cryogenic control system, based on the exploitation of the experience gained. These developments are always aimed at increasing the safety and improving the performance. To overcome the long-term maintenance challenges, key strategies such as the use of homogeneous hardware solutions and the optimization of the maintenance procedures were set up. They are easing the development of the control applications and the hardware configuration by allowing a structured and homogeneous approach. Furthermore, they reduce the needed manpower and minimize the financial impact of the periodical maintenance. In that context, the standardization of technical solutions both at hardware and software level simplify also the systems monitoring the operation and maintenance processes, while providing a high level of availability.

Usage Scenarios for Byte-Addressable Persistent Memory in High-Performance and Data Intensive Computing

Byte-addressable persistent memory (B-APM) presents a new opportunity to bridge the performance gap between main memory and storage. In this paper, we present the usage scenarios for this new technology, based on the capabilities of Intel's DCPMM. We outline some of the basic performance characteristics of DCPMM, and explain how it can be configured and used to address the needs of memory and I/O intensive applications in the HPC and data intensive domains. Two decision trees are presented to advise on the configuration options for B-APM; their use is illustrated with two examples. We show that the flexibility of the technology has the potential to be truly disruptive, not only because of the performance improvements it can deliver, but also because it allows systems to cater for wider range of applications on homogeneous hardware.

USCDRP: unequal secure cluster-based distributed routing protocol for wireless sensor networks

In the last decade, parallel computing techniques become more popular and offer various architectures to perform diverse processes in an individual computing node or a collection of strongly linked nodes with homogeneous hardware. In this paper, we propose a distributed secure unequal cluster-based multipath routing protocol (USCDRP) for improving the energy efficiency and also to enhance the security as well as reliability. The presented USCDRP is the enhanced work of SCMRP, which provides the security to a particular extent. The presented method offers reliability, energy efficiency and maximum lifetime. Besides, an independent clustering method is provided to eliminate the orphan nodes and cluster maintenance strategy to enhance the network lifetime. A detailed experimental analysis ensured the superiority of the USCDRP over the compared methods concerning throughput, energy efficiency and network lifetime analysis. The simulation outcome indicated that all the nodes in WSN become dead node, and the network becomes inactive in the round number of 2021 by Sec-LEACH, 2412 by SCMRP and 3245 by USCDRP.

Design of online control and monitoring software for the CPPF system in the CMS Level-1 trigger upgrade

The concentration preprocessing and fan-out (CPPF) system is one of the electronic subsystems of the upgraded Compact Muon Solenoid (CMS) Level-1 trigger system. It includes, in hardware, eight specially designed CPPF cards, one CMS card called AMC13, one commercial Micro-TCA Carrier HUB (MCH) card, and a Micro-TCA shelf. Powerful online software is needed for the system, including providing reliable configuration and monitoring for the hardware, and a graphical interface for executing all actions and publishing monitoring messages. Further, to control and monitor the large amount of homogeneous hardware, the SoftWare Automating conTrol of Common Hardware (SWATCH) concept was proposed and developed. The SWATCH provides a generic structure and is flexible for customization. The structure includes a hardware access library based on the IPbus protocol, which assumes a virtual 32-bit address/32-bit data bus and builds a simple hardware access layer. Furthermore, the structure provides a graphical user interface, which is based on modern web technology and is accessible by web page. The CPPF controlling and monitoring online software was also customized from a common SWATCH cell, and provides a finite state machine (FSM) for configuring the entire CPPF hardware, and five monitoring objects for periodically collecting monitoring data from five main functional modules in the CPPF hardware. This paper introduces the details of the CPPF SWATCH cell development.

Speedup and Power Scaling Models for Heterogeneous Many-Core Systems

Traditional speedup models, such as Amdahl’s law, Gustafson’s, and Sun and Ni’s, have helped the research community and industry better understand system performance capabilities and application parallelizability. As they mostly target homogeneous hardware platforms or limited forms of processor heterogeneity, these models do not cover newly emerging multi-core heterogeneous architectures. This paper reports on novel speedup and energy consumption models based on a more general representation of heterogeneity, referred to as the normal form heterogeneity, that supports a wide range of heterogeneous many-core architectures. The modelling method aims to predict system power efficiency and performance ranges, and facilitates research and development at the hardware and system software levels. The models were validated through extensive experimentation on the off-the-shelf big.LITTLE heterogeneous platform and a dual-GPU laptop, with an average error of 1 percent for speedup and of less than 6.5 percent for power dissipation. A quantitative efficiency analysis targeting the system load balancer on the Odroid XU3 platform was used to demonstrate the practical use of the method.

Is the Same Instance Type Created Equal? Exploiting Heterogeneity of Public Clouds

Public cloud platforms might start with homogeneous hardware; nevertheless, because of inevitable hardware upgrades, or adding more capacity, the initial homogeneous platform will gradually evolve into heterogeneous as time passes by. The consequent performance heterogeneity is of concern to cloud users. In this paper, we evaluate performance variations from hardware heterogeneity and scheduling mechanisms of public clouds. Amazon Elastic Compute Cloud (Amazon EC2) and Rackspace Cloud are used as the representatives because of their relatively long record and wide usage among small and medium enterprises (SMEs). A comprehensive set of microbenchmarks and application-level macrobenchmarks have been used to investigate performance variation. Several major contributions have been made. First, we find out that heterogeneous hardware is a commonality among the relatively long-lasting cloud platforms, although the level of heterogeneity varies. Second, we observe that heterogeneous hardware is the primary culprit of performance variation of cloud platforms. Third, we discover that varied CPU acquisition percentages and different virtual machine scheduling mechanisms exacerbate the performance variation problem, especially for network related operations. Finally, based on the observations, we propose cost-saving approaches and analyze Nash equilibrium from cloud user perspective. By using a simple "trial-and-better" approach, i.e., keep good-performing instances and discard bad-performing instances, cloud users can achieve up to 30 percent cost saving.

Scalable multi-GPU implementation of the MAGFLOW simulator

We have developed a robust and scalable multi-GPU (Graphics Processing Unit) version of the cellular-automaton-based MAGFLOW lava simulator. The cellular automaton is partitioned into strips that are assigned to different GPUs, with minimal overlapping. For each GPU, a host thread is launched to manage allocation, deallocation, data transfer and kernel launches; the main host thread coordinates all of the GPUs, to ensure temporal coherence and data integrity. The overlapping borders and maximum temporal step need to be exchanged among the GPUs at the beginning of every evolution of the cellular automaton; data transfers are asynchronous with respect to the computations, to cover the introduced overhead. It is not required to have GPUs of the same speed or capacity; the system runs flawlessly on homogeneous and heterogeneous hardware. The speed-up factor differs from that which is ideal (#GPUs×) only for a constant overhead loss of about 4E−2 · T · #GPUs, with T as the total simulation time.

Towards market-oriented Clouds

In the literature, Cloud computing is seen as an evolution of Grid computing, which itself is an evolution of distributed computing and supercomputing. The distinction between the Grid and Cloud concepts is not clearly defined, both of them referring to an externalization of the private data centres. Two main advantages are expected from these two approaches. Instead of investing in the installation and management of its own computing infrastructure, an enterprise may benefit from shared computing resources managed on a remote site by a third party. This third party is known as the Grid/Cloud service provider (GSP or CSP, respectively). This means that both capital expenditures and operational expenditures of the enterprise may be considerably reduced against a moderate increase in service expenditure. In the years 1990s, Grid computing was mainly developed in the perspective to provide affordable high-performance computing to small and medium enterprises. Circa 2000, Cloud Computing has been introduced by the service sector with the objective to extend the externalization of hardware resources to software resources. This distinction added to the fact that in 20 years, not only professionals but also residential users benefit from highspeed Internet access are at the origin of the Cloud computing concept. The economical models of Grid computing and of Cloud computing are quite different. Cloud computing enables the externalization of software resources at a very large scale for residential users; whereas; Grid computing mainly refers to the provisioning of externalized large computing facilities for professionals. Today, both types of services are accessible via simple Web portals. This evolution has motivated the emergence of numerous start-ups and companies specialized in service provisioning with an expected market size by 2011 around $95 billion in business and productivity applications (email, office, CRM, etc.) according to Merrill Lynch analysts. The fields of application of Cloud computing seems almost unlimited, all the sectors of our economy being concerned. Applications are ranging from distant medical diagnostic, collaborative image processing, scientific computation, financial operations, and industrial processes to radio astronomy already financially benefiting from advantages of Clouds or Grids. Two other criteria are also to be considered to distinguish between Grid and Cloud services. The consumers of Cloud services being in majority residential users, the mode of payment of the consumers is carried out on-the-fly with a duration-based charging. The consumers of Grid services being in majority enterprises, their first requirement is rather the availability of the computing resources than the price of the service itself. Resources are in that case reserved in advance for a longer period of time than in the case of Cloud services. A last distinction may be done between Grids and Clouds. In the case of Grids, heterogeneous computing resources managed by different entities may have to interoperate to provide the required service. This is not the case in the context of Clouds where comparatively, a smaller computing power is requested for a given service. Homogeneous hardware M. Gagnaire (*) Institut Telecom/Telecom ParisTech, Paris, France e-mail: gagnaire@telecom-paristech.fr

Dialogue Management: Support for Dialogue Independence

Dialogue management involves the application of systems in the design and delivery of user-system dialogues. This article describes a dialogue management environment that is based on a concept of dialogue independence: the separation of dialogue definition form dialogue execution. In this environment, software facilities provide building blocks for design and delivery of dialogues that are consistent, efficient, shareable, and portable. The environment supports multiple workstation and communication protocol environments. This environment is compared to those emerging in homogeneous hardware environments such as HyperCard.