High Performance Network of PC Cluster Maestro

This paper presents a design, an architecture, and performance evaluation of high-performance network of PC cluster, called Maestro. Most networks of recent clusters have been organized based on WAN or LAN technology, due to their market availability. However, communication protocols and functions of such conventional networks are not optimal for parallel computing, which requires low latency and high bandwidth communication. In this paper, we propose two optimizations for high-performance communication: (1) transferring in burst as many packets as the receiving buffer accepts at once, and (2) having each hardware component pass one data unit to another in a pipelined manner. We have developed a network interface and a switch, which are composed of dedicated hardware modules to realize these optimizations. An implementatin of the message passing library developed on Maestro cluster is also described. Performance evaluation shows that the proposed optimizations can extract the potential performance of the physical layer efficiently and improve the performance in communication.     

Effects of channel bandwidth variation on performance of individual messages in spatially and temporally heterogeneous communication networks

As the domain of communication systems grows, heterogeneity among computers and subnetworks employed for a task also increases. Channel bandwidth available for a message on a communication network varies with time and link. This variation can have a significant effect on performance of an individual message and also that of the network as a whole. Therefore, it is important to understand effects of bandwidth heterogeneity on the network performance in order to optimally utilize a heterogeneous communication network. The ability to use such a network optimally is highly desirable in many applications such as network-based data-intensive high performance computing. The main goal of this paper is to analyze effects of temporal and spatial heterogeneity on performance of individual messages in detail via an extensive simulation, in terms of throughput, end-to-end delay, etc. Also, the problems of path selection and multi-path data transfer are considered to illustrate how the analysis results may be used in the future effort of optimizing the network performance by taking channel bandwidth heterogeneity into account.     

心脏起搏器程控与遥测系统的研究

目的通过搭建一个心脏起搏器程控与遥测的实验平台,研究高效的程控遥测方案。方法系统由计算机显示终端、FPGA开发板、无线传输硬件电路构成。讨论除无线传输硬件电路之外的设计：计算机显示终端主要实现发送程控命令和显示遥测数据;FPGA开发板分别模仿程控仪和心脏起搏器,主要实现程控遥测信息的编码、解码和纠错功能。结论本系统提出了一种脉冲位置调制（PPM）编码方式——多位PPM编码方式,用于产生包含信息的脉冲波。该方法可以提高无线通讯的传输效率,若应用于植入式心脏起搏器,可延长其电池的使用寿命。实际测试结果证实了本方案的可行性。     

Memory retrieval as a p-adic dynamical system

We propose a mathematical model of the memory retrieval process based on dynamical systems over a metric space of p-adic numbers representing a configuration 'space of ideas' in which two ideas are close if they have a sufficiently long common root. Our aim is to suggest a new way of conceptualizing human memory retrieval that might be useful for simulation purposes or for the construction of artificial intelligence devices, as well as for a deeper understanding of the process itself. The dynamical system is assumed to be located in a blackbox processing unit (the 'subconscious') and controlled by an interface control unit (the 'conscious') that fixes parameters in the dynamical system and starts its iteration by sending an initial generating idea to it. We show that even simple p-adic dynamical systems admit behavioral scenarios that could explain some of the essential features of the human memory retrieval process.     

XenLoop: a transparent high performance inter-VM network loopback

Advances in virtualization technology have focused mainly on strengthening the isolation barrier between virtual machines (VMs) that are co-resident within a single physical machine. At the same time, a large category of communication intensive distributed applications and software components exist, such as web services, high performance grid applications, transaction processing, and graphics rendering, that often wish to communicate across this isolation barrier with other endpoints on co-resident VMs. State of the art inter-VM communication mechanisms do not adequately address the requirements of such applications. TCP/UDP based network communication tends to perform poorly when used between co-resident VMs, but has the advantage of being transparent to user applications. Other solutions exploit inter-domain shared memory mechanisms to improve communication latency and bandwidth, but require applications or user libraries to be rewritten against customized APIs—something not practical for a large majority of distributed applications. In this paper, we present the design and implementation of a fully transparent and high performance inter-VM network loopback channel, called XenLoop, in the Xen virtual machine environment. XenLoop does not sacrifice user-level transparency and yet achieves high communication performance between co-resident guest VMs. XenLoop intercepts outgoing network packets beneath the network layer and shepherds the packets destined to co-resident VMs through a high-speed inter-VM shared memory channel that bypasses the virtualized network interface. Guest VMs using XenLoop can migrate transparently across machines without disrupting ongoing network communications, and seamlessly switch between the standard network path and the XenLoop channel. In our evaluation using a number of unmodified benchmarks, we observe that XenLoop can reduce the inter-VM round trip latency by up to a factor of 5 and increase bandwidth by a up to a factor of 6.

DICOM-RT Ion interface to utilize MC simulations in routine clinical workflow for proton pencil beam radiotherapy

To adopt Monte Carlo (MC) simulations as an independent dose calculation method for proton pencil beam radiotherapy, an interface that converts the plan information in DICOM format into MC components such as geometries and beam source is a crucial element. For this purpose, a DICOM-RT Ion interface (https://github.com/topasmc/dicom-interface) has been developed and integrated into the TOPAS MC code to perform such conversions on-the-fly. DICOM-RT objects utilized in this interface include Ion Plan (RTIP), Ion Beams Treatment Record (RTIBTR), CT image, and Dose. Beamline geometries, gantry and patient coordinate systems, and fluence maps are determined from RTIP and/or RTIBTR. In this interface, DICOM information is processed and delivered to a MC engine in two steps. A MC model, which consists of beamline geometries and beam source, to represent a treatment machine is created by a DICOM parser of the interface. The complexities from different DICOM types, various beamline configurations and source models are handled in this step. Next, geometry information and beam source are transferred to TOPAS on-the-fly via the developed TOPAS extensions. This interface with two treatment machines was successfully deployed into our automated MC workflow which provides simulated dose and LET distributions in a patient or a water phantom automatically when a new plan is identified. The developed interface provides novel features such as handling multiple treatment systems based on different DICOM types, DICOM conversions on-the-fly, and flexible sampling methods that significantly reduce the burden of handling DICOM based plan or treatment record information for MC simulations.     

An Effective Logical Cache for a Clustered LRC-Based DSM System

As local-area workstation networks are widely available, the idea of offering a software distributed shared memory (SDSM) system across interconnects of clusters is quite an attractive alternative for compute-intensive applications. However, the higher cost of sending a message over an inter-cluster link compared to an intra-cluster one can limit applications' performance on a multi-cluster SDSM system. In this paper, we present the extensions that we have added to the SDSM TreadMarks, which provides the lazy release consistency (LRC) memory model, in order to adapt it to a loosely-coupled cluster-based platform. We have implemented a logical per-cluster cache that exploits cluster locality. By accessing the cache of its cluster, a processor can share data previously requested by a second processor of its cluster, thereby, minimizing, the cost of inter-cluster communication.     

The asymmetric self-assembly mechanism of adherens junctions: a cellular push-pull unit

To form adherens junctions (AJ), cells first establish contact by sending out lamellipodia onto neighboring cells. We investigated the role of contacting cells in AJ assembly by studying an asymmetric AJ motif: finger-like AJ extending across the cell-cell interface. Using a cytoskeleton replica and immunofluorescence, we observed that actin bundles embedded in the lamellipodia are co-localized with stress fibers in the neighboring cell at the AJ. This suggests that donor lamellipodia present actin fingers, which are stabilized by acceptor lamellae via acto-myosin contractility. Indeed, we show that changes in actin network geometry promoted by Rac overexpression lead to corresponding changes in AJ morphology. Moreover, contractility inhibition and enhancement (via drugs or local traction) lead respectively to the disappearance and further growth of AJ fingers. Thus, we propose that receiving lamellae exert a local pull on AJ, promoting further polymerization of the donor actin bundles. In spite of different compositions, AJ and focal contacts both act as cellular mechanosensors.     

Hardware implementation of a new artificial neuron

This paper describes an FPGA (Field Programmable Gate Arrays) implementation of a new type of neuron, the Quantron. The goal is to demonstrate the capability of current technology to closely recreate the human body's reaction to a change of temperature. This is accomplished by creating a function that adds a number of kernels at different frequencies depending on the external temperature. Once the sum of the kernels reaches a certain threshold, the artificial neural network, equivalent to its biological counterpart, "reacts" by sending a specific output signal designed to trigger a response. The various elements of each subsystem are discussed and implemented in software and hardware. The results are analyzed in terms of accuracy and efficiency compared to the biological equivalent.     

Efficient and robust associative memory from a generalized Bloom filter

We develop a variant of a Bloom filter that is robust to hardware failure and show how it can be used as an efficient associative memory. We define a measure of the information recall and show that our new associative memory is able to recall more than twice as much information as a Hopfield network. The extra efficiency of our associative memory is all the more remarkable as it uses only bits while the Hopfield network uses integers.     

Towards an efficient parallel raycasting of unstructured volumetric data on distributed environments

Direct volume rendering of large and unstructured datasets demands high computational power and memory bandwidth. Developing an efficient parallel algorithm requires a deep understanding of the bottlenecks involved in the solutions for this problem. In this work, we make a thorough analysis of the overhead components involved in parallel volume raycasting of unstructured grids for high-resolution images on distributed environments. This evaluation has revealed potential opportunities for performance improvements. The result is a novel approach to distributed memory raycasting that includes different acceleration techniques to enhance ray distribution, face projection, memory locality, and message exchanging, while maintaining load balance. We report the gains achieved in each phase and in the complete parallel algorithm when compared with a conventional approach.     

Inhibition of adhesion of F9 embryonal carcinoma cells to substratum by a novel monoclonal antibody, TEC-05, reactive with a developmentally regulated carbohydrate epitope

Embryonal carcinoma cells carry on their surfaces carbohydrate antigens that are also expressed in early embryonic cells. We report here the expression and properties of a new developmentally regulated carbohydrate epitope, which is defined by a monoclonal antibody TEC-05. This antibody was generated by immunization of a rat with mouse embryonal carcinoma cells P19S1801A1. By immunofluorescence, the TEC-5 epitope was first detected on 8-cell-stage mouse embryos and was present on all subsequent stages of preimplantation development. Absorption analysis revealed that TEC-5 epitope was expressed only on a limited number of adult mouse tissues. In the direct radioantibody binding assay, TEC-05 reacted strongly with OTF9-63 cells and with some of the mouse embryonal carcinoma cell lines tested. Its reaction with differentiated cell lines was weak or undetectable. In the course of differentiation of OTF9-63 cells induced by retinoic acid, the epitope disappeared with the onset of morphological differentiation. The binding of the antibody to OTF9-63 cells was inhibited to 50% by 10-50 microM N-acetyllactosamine and lactose. Immunolabelling of extracts from OTF9-63 cells separated by sodium-dodecyl-sulfate (SDS) polyacrylamide gel electrophoresis revealed that TEC-5 epitope was carried by high-molecular-weight glycoconjugates (molecular weight greater than 100,000). Molecules, isolated from [3H]-fucose-labelled OTF9-63 cells by indirect immunoprecipitation with TEC-05 antibody, were degraded by extensive pronase digestion or mild alkaline treatment to large carbohydrate chains that were excluded from a Sephadex G-50 column. Direct evidence that TEC-05 antibody bound to embryoglycan was obtained using a modified Farr's assay. The antibody was found to inhibit adhesion of F9 and OTF9-63 cells to substratum. The inhibitory effect, which could be abrogated by lactose, seemed to be specific, because another IgM monoclonal antibody which also binds to embryoglycan had no effect. Combined data indicated that TEC-05 antibody recognizes a carbohydrate epitope which is involved in cell-substratum adhesion of F9 cells and which provides a new marker for structure-function studies of stage-specific embryonic antigens.     

Conjunctive spatial and self-motion codes are topographically organized in the GABAergic cells of the lateral septum

The hippocampal spatial code’s relevance for downstream neuronal populations—particularly its major subcortical output the lateral septum (LS)—is still poorly understood. Here, using calcium imaging combined with unbiased analytical methods, we functionally characterized and compared the spatial tuning of LS GABAergic cells to those of dorsal CA3 and CA1 cells. We identified a significant number of LS cells that are modulated by place, speed, acceleration, and direction, as well as conjunctions of these properties, directly comparable to hippocampal CA1 and CA3 spatially modulated cells. Interestingly, Bayesian decoding of position based on LS spatial cells reflected the animal’s location as accurately as decoding using the activity of hippocampal pyramidal cells. A portion of LS cells showed stable spatial codes over the course of multiple days, potentially reflecting long-term episodic memory. The distributions of cells exhibiting these properties formed gradients along the anterior–posterior and dorsal–ventral axes of the LS, directly reflecting the topographical organization of hippocampal inputs to the LS. Finally, we show using transsynaptic tracing that LS neurons receiving CA3 and CA1 excitatory input send projections to the hypothalamus and medial septum, regions that are not targeted directly by principal cells of the dorsal hippocampus. Together, our findings demonstrate that the LS accurately and robustly represents spatial, directional as well as self-motion information and is uniquely positioned to relay this information from the hippocampus to its downstream regions, thus occupying a key position within a distributed spatial memory network.

Calcium imaging of neurons in freely behaving mice reveals how the lateral septum, the main output of the hippocampal place cells, effectively represents information about not only location, but also head direction and self-movement, and may be pivotal in sending this information to downstream brain regions.     

Task offloading method of edge computing in internet of vehicles based on deep reinforcement learning

Compared with the traditional network tasks, the emerging Internet of Vehicles (IoV) technology has higher requirements for network bandwidth and delay. However, due to the limitation of computing resources and battery capacity of existing mobile devices, it is hard to meet the above requirements. How to complete task offloading and calculation with lower task delay and lower energy consumption is the most important issue. Aiming at the task offloading system of the IoV, this paper considers the situation of multiple MEC servers when modeling, and proposes a dynamic task offloading scheme based on deep reinforcement learning. It improves the traditional Q-Learning algorithm and combines deep learning with reinforcement learning to avoid dimensional disaster in the Q-Learning algorithm. Simulation results show that the proposed algorithm has better performance on delay, energy consumption, and total system overhead under the different number of tasks and wireless channel bandwidth.

     

Discrimination of jittered sonar echoes by the echolocating bat,Eptesicus fuscus: The shape of target images in echolocation

1. Behavioral experiments with jittering echoes examined acoustic images of sonar targets in the echolocating bat, Eptesicus fuscus, along the echo delay or target range axis. Echo phase, amplitude, bandwidth, and signal-to-noise ratio were manipulated to assess the underlying auditory processes for image formation. 2. Fine delay acuity is about 10 ns. Calibration and control procedures indicate that this represents temporal acuity rather than spectral discrimination. Jitter discrimination curves change in phase when the phase of one jittering echo is shifted by 180 degrees relative to the other, showing that echo phase is involved in delay estimation. At an echo detectability index of about 36 dB, fine acuity is 40 ns, which is approximately as predicted for the delay accuracy of an ideal receiver. 3. Compound performance curves for 0 degrees and 180 degrees phase conditions match the crosscorrelation function of the echoes. The locations of both 0 degrees and 180 degrees phase peaks in the performance curves shift along the time axis by an amount that matches neural amplitude-latency trading in Eptesicus, confirming a temporal basis for jitter discrimination.     

MATIN: A Random Network Coding Based Framework for High Quality Peer-to-Peer Live Video Streaming

In recent years, Random Network Coding (RNC) has emerged as a promising solution for efficient Peer-to-Peer (P2P) video multicasting over the Internet. This probably refers to this fact that RNC noticeably increases the error resiliency and throughput of the network. However, high transmission overhead arising from sending large coefficients vector as header has been the most important challenge of the RNC. Moreover, due to employing the Gauss-Jordan elimination method, considerable computational complexity can be imposed on peers in decoding the encoded blocks and checking linear dependency among the coefficients vectors. In order to address these challenges, this study introduces MATIN which is a random network coding based framework for efficient P2P video streaming. The MATIN includes a novel coefficients matrix generation method so that there is no linear dependency in the generated coefficients matrix. Using the proposed framework, each peer encapsulates one instead of n coefficients entries into the generated encoded packet which results in very low transmission overhead. It is also possible to obtain the inverted coefficients matrix using a bit number of simple arithmetic operations. In this regard, peers sustain very low computational complexities. As a result, the MATIN permits random network coding to be more efficient in P2P video streaming systems. The results obtained from simulation using OMNET++ show that it substantially outperforms the RNC which uses the Gauss-Jordan elimination method by providing better video quality on peers in terms of the four important performance metrics including video distortion, dependency distortion, End-to-End delay and Initial Startup delay.     

Design and analysis of a distributed grid resource discovery protocol

Computational grids have been emerging as a new paradigm for solving large complex problems over the recent years. The problem space and data set are divided into smaller pieces that are processed in parallel over the grid network and reassembled upon completion. Typically, resources are logged into a resource broker that is somewhat aware of all of the participants available on the grid. The resource broker scheme can be a bottleneck because of the amount of computational power and network bandwidth needed to maintain a fresh view of the grid. In this paper, we propose to place the load of managing the network resource discovery on to the network itself: inside of the routers. In the proposed protocol, the routers contain tables for resources similar to routing tables. These resource tables map IP addresses to the available computing resource values, which are provided through a scoring mechanism. Each resource provider is scored based on the attributes they provide such as the number of processors, processor frequency, amount of memory, hard drive space, and the network bandwidth. The resources are discovered on the grid by the protocol’s discovery packets, which are encapsulated within the TCP/IP packets. The discovery packet visits the routers and look up in the resource tables until a satisfactory resource is found. The protocol is validated by simulations with five different deployment environments.     

Approximate quantum trajectory dynamics for reactive processes in condensed phase

A method of molecular dynamics with quantum corrections, practical for studies of large molecular systems, is reviewed. The approach is based on the Bohmian formulation of the time-dependent Schrödinger equation in which a wavefunction is represented by an ensemble of interdependent trajectories. The quantum effects come from the quantum potential acting on trajectories on par with the usual classical potential. The quantum potential is determined from the evolving nuclear wavefunction, i.e. from the quantum trajectory (QT) ensemble itself. For practical and conceptual reasons the quantum potential and corresponding quantum nuclear effect are computed only for the selected light nuclei. For studies of reactive chemical processes, the classical potential is computed on-the-fly using the density functional tight binding method of electronic structure. A massively parallel implementation, based on the message passing interface allows for efficient simulations of ensembles of thousands of trajectories describing systems of up to 200 atoms. As a biochemical application, the approximate QT approach is used to model the tunnelling-dominated proton transfer in soybean-lipoxygenase-1. A materials science application is represented by a study of the nuclear quantum effect on adsorption of hydrogen and deuterium on a C₃₇H₁₅ molecule, which is a model ‘flake’ of graphene.