Local Workstation Research Articles

Running high resolution coastal models in forecast systems: Moving from workstations and HPC cluster to cloud resources

Computational forecast systems (CFS) are essential modelling tools for coastal management by providing water dynamics predictions. Nowadays CFS are processed in dedicated workstations, fulfilling quality control through automatic comparison with field data. Recently, CFS has been successfully ported to High Performance Computing (HPC) resources, maintained by highly-specialized staff in these complex environments. The need to increase the available resources for more demanding applications and to enhance the portability for use in non-scientific institutions has promoted the search for more flexible and user-friendly approaches. The scalability and flexibility of cloud resources, with dedicated services for facilitating their use, makes them an attractive option.Herein, the performance of CFS using ECO-SELFE MPI-based model is assessed and compared for the first time in multiple environments, including local workstations, an HPC cluster and a pilot cloud. The analysis is conducted in a range of resources from the physical core count available at the smaller resources to the optimal number of processes, using cloud and HPC cluster resources. Results for the smaller, common physical resources show that the cloud is an attractive option for CFS operation. As the optimal number of processes for the use case is at the limit of the workstations common pool, an analysis was also performed using HPC cluster nodes and federated MPI resources. Results show that the cloud remains an attractive option for CFS. This conclusion is valid both for the use of a single host or through federated hosts, providing that efficient communication infrastructure (such as SRIOV) is available.

Using the Launcher for Executing High Throughput Workloads

For many scientific disciplines, the transition to using advanced cyberinfrastructure comes not out of a desire to use the most advanced or most powerful resources available, but because their current operational model is no longer sufficient to meet their computational needs. Many researchers begin their computations on their desktop or local workstation, only to discover that the time required to simulate their problem, analyze their instrument data, or score the multitude of entities that they want to would require far more time than they have available. Launcher is a simple utility which enables the execution of high throughput computing workloads on managed HPC systems quickly and with as little effort as possible on the part of the user. Basic usage of the Launcher is straightforward, but Launcher provides several more advanced capabilities including use of Intel® Xeon Phi™ coprocessor cards and task binding support for multi-/many-core architectures. We step through the processes of setting up a basic Launcher job, including creating a job file, setting appropriate environment variables, and using scheduler integration. We also describe how to enable use of the Intel® Xeon Phi™ coprocessor cards, take advantage of Launcher's task binding system, and execute many parallel (OpenMP/MPI) applications at once.

Aligning, analyzing, and visualizing sequences for antibody engineering: Automated recognition of immunoglobulin variable region features.

The analysis and comparison of large numbers of immunoglobulin (Ig) sequences that arise during an antibody selection campaign can be time-consuming and tedious. Typically, the identification and annotation of framework as well as complementarity-determining regions (CDRs) is based on multiple sequence alignments using standardized numbering schemes, which allow identification of equivalent residues among different family members but often necessitate expert knowledge and manual intervention. Moreover, due to the enormous length variability of some CDRs the benefit of conventional Ig numbering schemes is limited and the calculation of correct sequence alignments can become challenging. Whereas, in principle, a well established set of rules permits the assignment of CDRs from the amino acid sequence alone, no currently available sequence alignment editor provides an algorithm to annotate new Ig sequences accordingly. Here we present a unique pattern matching method implemented into our recently developed ANTICALIgN editor that automatically identifies all hypervariable and framework regions in experimentally elucidated antibody sequences using so-called "regular expressions." By combination of this widely supported software syntax with the unique capabilities of real-time aligning, editing and analyzing extended sets of amino acid and/or nucleotide sequences simultaneously on a local workstation, ANTICALIgN provides a powerful utility for antibody engineering. Proteins 2016; 85:65-71. © 2016 Wiley Periodicals, Inc.

Easy Access to HPC Resources through the Application GUI

The computing environment at the King Abdullah University of Science and Technology (KAUST) is growing in size and complexity. KAUST hosts the tenth fastest supercomputer in the world (Shaheen II) and several HPC clusters. Researchers can be inhibited by the complexity, as they need to learn new languages and execute many tasks in order to access the HPC clusters and the supercomputer. In order to simplify the access, we have developed an interface between the applications and the clusters and supercomputer that automates the transfer of input data and job submission and also the retrieval of results to the researcher’s local workstation. The innovation is that the user now submits his jobs from within the application GUI on his workstation, and does not have to directly log into the clusters or supercomputer anymore. This article details the solution and its benefits to the researchers.

On the Virtualization of CUDA Based GPU Remoting on ARM and X86 Machines in the GVirtuS Framework

The astonishing development of diverse and different hardware platforms is twofold: on one side, the challenge for the exascale performance for big data processing and management; on the other side, the mobile and embedded devices for data collection and human machine interaction. This drove to a highly hierarchical evolution of programming models. GVirtuS is the general virtualization system developed in 2009 and firstly introduced in 2010 enabling a completely transparent layer among GPUs and VMs. This paper shows the latest achievements and developments of GVirtuS, now supporting CUDA 6.5, memory management and scheduling. Thanks to the new and improved remoting capabilities, GVirtus now enables GPU sharing among physical and virtual machines based on x86 and ARM CPUs on local workstations, computing clusters and distributed cloud appliances.

96-69: Rapid estimation of 3D ventricular activation from electroanatomic mapping

Purpose Heart failure patients are often affected by a ventricular conduction disorder. Accurate characterization of the tridimensional (3D) activation sequence of the ventricles is fundamental for an effective personalized therapy. Electroanatomical mapping provides the activation times at the endocardium, with limited coverage due to anatomical constraints. Our aim is to develop a robust tool to generate a complete 3D, biventricular activation map from a limited number of endocardial measurements. Method In 6 candidates to cardiac resynchronization therapy a high density endocardial activation map of the right and left ventricle was obtained. High resolution propagating activation was simulated in patient-tailored anatomical heart models using the anisotropic eikonal equation and realistic fiber orientation. Subsequently, the mean square error between the measured and simulated activation times was minimized by varying early activation site(s) and local anisotropic conduction velocities. Results. A close correlation was achieved between the simulated and measured activation time for a single patient (Panel A) while detailed 3D activation maps were achieved (panel B and C). In all cases, the optimal parameters were identified within few minutes and the full activation sequence was computed in less than 2 seconds. A 0.9 to 0.95 Pearson correlation between measured and simulated activation times was obtained in all cases. The mathematical formulation is flexible in terms of the number of activation sites and a possible delay between them, in order to mimic the presence of a partially working Purkinje network. Conclusion Our methodology is able to generate in a clinically affordable computational time (few minutes) reliable, patient-specific, fully volumetric activation map of both ventricles using a local workstation or a laptop.

An automated, open-source pipeline for mass production of digital elevation models (DEMs) from very-high-resolution commercial stereo satellite imagery

We adapted the automated, open source NASA Ames Stereo Pipeline (ASP) to generate digital elevation models (DEMs) and orthoimages from very-high-resolution (VHR) commercial imagery of the Earth. These modifications include support for rigorous and rational polynomial coefficient (RPC) sensor models, sensor geometry correction, bundle adjustment, point cloud co-registration, and significant improvements to the ASP code base. We outline a processing workflow for ∼0.5m ground sample distance (GSD) DigitalGlobe WorldView-1 and WorldView-2 along-track stereo image data, with an overview of ASP capabilities, an evaluation of ASP correlator options, benchmark test results, and two case studies of DEM accuracy. Output DEM products are posted at ∼2m with direct geolocation accuracy of <5.0m CE90/LE90. An automated iterative closest-point (ICP) co-registration tool reduces absolute vertical and horizontal error to <0.5m where appropriate ground-control data are available, with observed standard deviation of ∼0.1–0.5m for overlapping, co-registered DEMs (n=14,17). While ASP can be used to process individual stereo pairs on a local workstation, the methods presented here were developed for large-scale batch processing in a high-performance computing environment. We are leveraging these resources to produce dense time series and regional mosaics for the Earth’s polar regions.

Manycast Overlay in Split-Incapable Networks for Supporting Bandwidth-Intensive Applications

Recent trends in science applications call for long-range and large-scale collaboration among laboratories and super-computing sites. Long gone are the days of entering data manually into a spreadsheet on a local workstation. The world's most powerful and ground-breaking experiments generate exabytes of information, which must be distributed to multiple labs for analysis and interpretation. Such trends reveal the unwavering importance of new communication paradigms, like multicasting and manycasting, which provide point-to-multipoint data transfers. Typically, these all-important mechanisms are provided at the optical layer, where split-capable cross-connects split input signals into multiple output signals all-optically. Unfortunately, some of the world's largest and most powerful networks do not have the hardware infrastructure to support such functionality, but allow for point-to-point communication exclusively. In such split-incapable (SI) networks, multicast and manycast must be provided as a logical overlay to the pre-existing and limited unicast infrastructure. In this paper, we present two overlay models for providing manycast support in SI networks: Manycasting with Drop at Member Node (MA-DMN) and Manycasting with Drop at Any Node (MA-DAN). Through the development of integer linear programs (ILPs) and heuristics, we evaluate these models in terms of both optimal solutions and efficient approximations for both small-scale and large-scale networks and consider both static and dynamic traffic scenarios. Our results demonstrate that despite a small tradeoff in additional complexity and delay from signal conversion to the optical domain, our models provide efficient utilization of network resources and greatly surpass the standard naive approach of establishing paths to every destination.

Interactive Access to LP DAAC Satellite Data Archives Through a Combination of Open-Source and Custom Middleware Web Services

Abstract-Current methods of searching for and retrieving data from satellite land remote sensing archives do not allow for interactive information extraction. Instead, Earth science data users are required to download files over low-bandwidth networks to local workstations and process data before science questions can be addressed.

A Web-Based Image Processing and Plan Evaluation Platform (WIPPEP) for Future Cloud-Based Radiation Therapy

Visualization and processing of medical images and radiation treatment plan evaluation have traditionally been constrained to local workstations with limited computation power and ability of data sharing and software update. We present a web-based image processing and planning evaluation platform (WIPPEP) for radiation therapy applications with high efficiency, ubiquitous web access, and real-time data sharing.

A Study of the Design of Wireless Medical Sensor Network based u-Healthcare System

Wireless Medical Sensor Networks and Radio Frequency Identification has been widely adopted in both medical and healthcare industry. This paper deals with the study of a WMSN and RFID based u-Healthcare system. An efficient framework for a u-Healthcare system utilizing RFID and WMSN has been presented. The system is capable of monitoring the patient’s medical status by using RFID body sensor and wirelessly transmits the medical data to a local workstation (WMSN gateway) before transmitting it to the central database server. Due to the patient’s movements, WMSN node’s movements will be patterned with the functionality of the Mobile IPv6.

Using ssh and sshfs to virtualize Grid job submission with RCondor

The HTCondor based glideinWMS has become the product of choice for exploiting Grid resources for many communities. Unfortunately, its default operational model expects users to log into a machine running a HTCondor schedd before being able to submit their jobs. Many users would instead prefer to use their local workstation for everything. A product that addresses this problem is RCondor, a module delivered with the HTCondor package. RCondor provides command line tools that simulate the behavior of a local HTCondor installation, while using ssh under the hood to execute commands on the remote node instead. RCondor also interfaces with sshfs, virtualizing access to remote files, thus giving the user the impression of a truly local HTCondor installation. This paper presents a detailed description of RCondor, as well as comparing it to the other methods currently available for accessing remote HTCondor schedds.

The Development of the Intelligent Anti-Disoperation Operation Order System

The intelligent anti-disoperation operation order system is developed. The system is established on operation management system (OMS). OMS provides the underlying technology support for operation order system. The supports of OMS include basic resources such as graphics, device, topology, the integration of real-time data sources, and the interfaces with external systems and so on. The full process management of operation order as well as operation security checking function based on automation system are realized. The system is composed of the relevant servers of OMS, communication server of local dispatch center, dispatcher workstations and other workstations, user workstations of substation, etc. The system is graphics, intelligent application system for dispatcher and supervisor. It is mainly used to many functions such as intelligent operation order generating, multilayered security checking of operating instructions, operation order management in accordance with procedure, offline operation order generating, operation training, etc.

SU-D-BRD-02: A Web-Based Image Processing and Plan Evaluation Platform (WIPPEP) for Future Cloud-Based Radiotherapy

Purpose: Visualization and processing of medical images and radiation treatment plan evaluation have traditionally been constrained to local workstations with limited computation power and ability of data sharing and software update. We present a web-based image processing and planning evaluation platform (WIPPEP) for radiotherapy applications with high efficiency, ubiquitous web access, and real-time data sharing. Methods: This software platform consists of three parts: web server, image server and computation server. Each independent server communicates with each other through HTTP requests. The web server is the key component that provides visualizations and user interface through front-end web browsers and relay information to the backend to process user requests. The image server serves as a PACS system. The computation server performs the actual image processing and dose calculation. The web server backend is developed using Java Servlets and the frontend is developed using HTML5, Javascript, and jQuery. The image server is based on open source DCME4CHEE PACS system. The computation server can be written in any programming language as long as it can send/receive HTTP requests. Our computation server was implemented in Delphi, Python and PHP, which can process data directly or via a C++ program DLL. Results: This software platform is running on a 32-core CPU server virtually hosting the web server, image server, and computation servers separately. Users can visit our internal website with Chrome browser, select a specific patient, visualize image and RT structures belonging to this patient and perform image segmentation running Delphi computation server and Monte Carlo dose calculation on Python or PHP computation server. Conclusion: We have developed a webbased image processing and plan evaluation platform prototype for radiotherapy. This system has clearly demonstrated the feasibility of performing image processing and plan evaluation platform through a web browser and exhibited potential for future cloud based radiotherapy.

Making distributed ALICE analysis simple using the GRID plug-in

We have developed an interface within the ALICE analysis framework that allows transparent usage of the experiment's distributed resources. This analysis plug-in makes it possible to configure back-end specific parameters from a single interface and to run with no change the same custom user analysis in many computing environments, from local workstations to PROOF clusters or GRID resources. The tool is used now extensively in the ALICE collaboration for both end-user analysis and large scale productions.

Interventional spinal procedures guided and controlled by a 3D rotational angiographic unit

The aim of this paper is to demonstrate the usefulness of 2D multiplanar reformatting images (MPR) obtained from rotational acquisitions with cone-beam computed tomography technology during percutaneous extra-vascular spinal procedures performed in the angiography suite. We used a 3D rotational angiographic unit with a flat panel detector. MPR images were obtained from a rotational acquisition of 8s (240 images at 30fps), tube rotation of 180° and after post-processing of 5s by a local work-station. Multislice CT (MSCT) is the best guidance system for spinal approaches permitting direct tomographic visualization of each spinal structure. Many operators, however, are trained with fluoroscopy, it is less expensive, allows real-time guidance, and in many centers the angiography suite is more frequently available for percutaneous procedures. We present our 6-year experience in fluoroscopy-guided spinal procedures, which were performed under different conditions using MPR images. We illustrate cases of vertebroplasty, epidural injections, selective foraminal nerve root block, facet block, percutaneous treatment of disc herniation and spine biopsy, all performed with the help of MPR images for guidance and control in the event of difficult or anatomically complex access. The integrated use of "CT-like" MPR images allows the execution of spinal procedures under fluoroscopy guidance alone in all cases of dorso-lumbar access, with evident limitation of risks and complications, and without need for recourse to MSCT guidance, thus eliminating CT-room time (often bearing high diagnostic charges), and avoiding organizational problems for procedures that need, for example, combined use of a C-arm in the CT room.

SU‐E‐E‐05: The Compute Cloud, a Massive Computing Resource for Patient‐Independent Monte Carlo Dose Calculations and Other Medical Physics Applications

Purpose: Massive computing resources are needed to generate an extensive linac phase‐space (PHSP) database for different radiotherapy beams that could serve as a starting point for several user‐applications, such as patient‐dependent dose calculations or Monte‐Carlo‐based quality assurance procedures. The goal of this project is to disseminate our knowledge about performing patient‐independent Monte‐Carlo dose calculations on the Amazon Elastic Compute Cloud (EC2) and educate medical physicists on technical aspects such as simulation instance management and cluster coordination while ensuring data security for a low CPU cost. Methods: Amazon‐Web‐Services, along with Google, Microsoft, IBM, etc, offers resizable compute capacity in the cloud changing the economics of computing by eliminating the server fixed expenses. A cost‐efficient elastic cluster is available for a wide users community who can perform Monte Carlo simulations or other numerical applications using Amazon Machine Images. These virtual machines can be configured to work in a particular scientific environment, i.e. GEANT4 or PENELOPE. GEANT4 patient‐independent dose simulations were launched in parallel on EC2 using high CPU extra‐large spot instances (7 GB of memory, 8 virtual cores at 2.13 to 2.44GHz). Data storage was performed using the Simple Storage Service (S3). The S3 Organizer was used to upload the simulation or download multiple files. The 4‐layer EC2 security system, composed of the host/guest operating systems, Firewall, and signed secret access keys, ensures data privacy. Results: A local script was set‐up to launch multiple dose calculation instances on EC2 and automatically save the results on S3. Geant4‐application development, testing, and debugging were performed on a local workstation, along with data analysis and comparison with experiment. IAEA‐compliant patient‐independent PHSP files were generated and dose calculations were successfully validated against experiment. Conclusions: Cloud computing offers a low‐cost alternative for improved radiotherapy planning and optimization algorithms by enabling beam transport and Monte Carlo‐based dose calculations.Work supported by Varian Medical Systems

20: Coupling of morphological and functional imaging on a X-rays fluorescence bimodal tomographic system for pre-clinical studies

We adapted the automated, open source NASA Ames Stereo Pipeline (ASP) to generate digital elevation models (DEMs) and orthoimages from very-high-resolution (VHR) commercial imagery of the Earth. These modifications include support for rigorous and rational polynomial coefficient (RPC) sensor models, sensor geometry correction, bundle adjustment, point cloud co-registration, and significant improvements to the ASP code base. We outline a processing workflow for ∼0.5 m ground sample distance (GSD) DigitalGlobe WorldView-1 and WorldView-2 along-track stereo image data, with an overview of ASP capabilities, an evaluation of ASP correlator options, benchmark test results, and two case studies of DEM accuracy. Output DEM products are posted at ∼2 m with direct geolocation accuracy of <5.0 m CE90/LE90. An automated iterative closest-point (ICP) co-registration tool reduces absolute vertical and horizontal error to <0.5 m where appropriate ground-control data are available, with observed standard deviation of ∼0.1–0.5 m for overlapping, co-registered DEMs (n = 14, 17). While ASP can be used to process individual stereo pairs on a local workstation, the methods presented here were developed for large-scale batch processing in a high-performance computing environment. We are leveraging these resources to produce dense time series and regional mosaics for the Earth’s polar regions.

How to Protect Patients Digital Images/Thermograms Stored on a Local Workstation

To ensure the security and privacy of patient electronic medical information stored on local workstations in doctors’ offices, clinic centers, etc., it is necessary to implement a secure and reliable method for logging on and accessing this information. Biometrically-based identification technologies use measurable personal properties (physiological or behavioral) such as a fingerprint in order to identify or verify a person’s identity, and provide the foundation for highly secure personal identification, verification and/or authentication solutions. The use of biometric devices (fingerprint readers) is an easy and secure way to log on to the system. We have provided practical tests on HP notebooks that have the fingerprint reader integrated. Successful/failed logons have been monitored and analyzed, and calculations have been made. This paper presents the false rejection rates, false acceptance rates and failure to acquire rates.

Trusting third-party storage providers for holding personal information. A context-based approach to protect identity-related data in untrusted domains

The never ending growth of digital information and the availability of low-cost storage facilities and networks capacity is leading users towards moving their data to remote storage resources. Since users’ data often holds identity-related information, several privacy issues arise when data can be stored in untrusted domains. In addition digital identity management is becoming extremely complicated due to the identity replicas proliferation necessary to get authentication in different domains. GMail and Amazon Web Services, for instance, are two examples of online services adopted by million of users throughout the world which hold huge amounts of sensitive users data. State-of-the-art encryption tools for large-scale distributed infrastructures allow users to encrypt content locally before storing it on a remote untrusted repository. This approach can experience performance drawbacks, when very large data-sets must be encrypted/decrypted on a single machine. The proposed approach extends the existing solutions by providing two additional features: (1) the encryption can also be delegated to a pool of remote trusted computing resources, and (2) the definition of the encryption context which drives the tool to select the best strategy to process the data. The performance benchmarks are based on the results of tests carried out both on a local workstation and on the Grid INFN Laboratory for Dissemination Activities (GILDA) testbed.