Using node and socket information to implement MPI Cartesian topologies

Computation/communication overlap is one of the fundamental techniques in hiding communication latency. Independent progress support in messaging layer, network interface offload capability and application usage of non-blocking communications are believed to increase overlap and yield performance benefits. In this paper, we analyze four MPI implementations on top of three high-speed interconnects (InfiniBand, Myrinet and iWARP Ethernet) in their ability to support overlap and communication progress. The results confirm that the offload ability needs to be supported with communication progress to increase the level of overlap. Our progress engine micro-benchmark results show that in all examined networks transferring small messages makes an acceptable level of progress and overlap. On the other hand, in most cases, transferring large messages does not make progress independently, decreasing the chances of overlap in applications.

GridFTP, designed using the Globus XIO framework, is one of the most popular methods in use to perform data transfers in the grid environment. But the performance of GridFTP in WAN is limited by the relatively low communication bandwidth offered by the existing network protocols. On the other hand, modern interconnects such as InfiniBand, with many advanced communication features like zero-copy protocol and RDMA operations, can greatly improve communication efficiency. In this paper, we take on the challenge of combining the ease of use of the Globus XIO framework and the high performance achieved through InfiniBand communication, thereby natively sup-porting GridFTP over InfiniBand based networks. The Advanced Data Transfer Service (ADTS), designed in our previous work, provides the low level InfiniBand support to the Globus XIO layer. We introduce the concepts of I/Ostaging in the Globus XIO ADTS driver to achieve efficient disk based data transfers. We evaluate our designs in both LAN and WAN environments using micro benchmarks as well as communication traces from several real world applications. We also provide insights into the communication performance with some in-depth analysis. Our experimental evaluation shows a performance improvement of up to 100% for ADTS based data transfers as opposed to TCP or UDP based ones in LAN and high delay WAN scenarios.

Looking at the TOP 500 list of supercomputers we can see that different architectures and networking technologies appear on the scene from time to time. The networking technologies are also changing along with the advances of processor technologies. While the hardware has been constantly changing, parallel applications written in different paradigms have remained largely unchanged. With MPI being the most popular parallel computing standard, it is crucial to have an MPI implementation portable across different networks and architectures. It is also desirable to have such an MPI deliver high performance. In this paper we take on this challenge. We have designed an MPI with both portability and portable high performance using the emerging uDAPL interface. We present the design alternatives and a comprehensive performance evaluation of this new design. The results show that this design can improve the startup time and communication performance by 30% compared with our previous work. It also delivers the same good performance as MPI implemented over native APIs of the underlying interconnect. We also present a multistream MPI design which aims to achieve high bandwidth across networks and operating systems. Experimental results on Solaris show that the multi-stream design can improve bandwidth over InfiniBand by 30%, and improve the application performance by up to 11%.

MPI / OpenMP mixed mode codes could potentially offer the most effective parallelisation strategy for an SMP cluster, as well as allowing the different characteristics of both paradigms to be exploited to give the best performance on a single SMP. This paper discusses the implementation, development and performance of mixed mode MPI / OpenMP applications. The results demonstrate that this style of programming will not always be the most effective mechanism on SMP systems and cannot be regarded as the ideal programming model for all codes. In some situations, however, significant benefit may be obtained from a mixed mode implementation. For example, benefit may be obtained if the parallel (MPI) code suffers from: poor scaling with MPI processes due to load imbalance or too fine a grain problem size, memory limitations due to the use of a replicated data strategy, or a restriction on the number of MPI processes combinations. In addition, if the system has a poorly optimised or limited scaling MPI implementation then a mixed mode code may increase the code performance.

This paper demonstrates how the principles of co-design, the merging of communication technology with control theory, can simultaneously improve both control and communication performance within an IEEE 802.11b wireless networked control system (WNCS). Data rate scaling is first described, which is a special property within 802.11 wireless networks that improves the frame error ratio (FER) of the communication channel under poor channel conditions. This is followed by a description of sample rate adaptation, a control technique that adapts the sampling interval in response to time-varying network parameters. Although both schemes are effective individually, significant improvement in performance was seen when both were used together, thereby highlighting the benefits of co-design. Supportive results are presented from a simulated cart-mounted inverted pendulum system.

In this paper, we present a new thread-safe implementation of MPI, called MiMPI, that allows efficient development of thread-safe parallel programs using the message-passing standard MPI. MiMPI uses multithread operations to increase the performance of collective communications. The paper describes the main design goals of MiMPI and the current implementation. Finally we present some performance results, obtained on an IBM SP2, comparing the performance with others MPI implementations.

Many scientific applications perform computations on a Cartesian grid. The common approach for the parallelization of these applications with MPI is domain decomposition. To help developers with the mapping of MPI processes to subdomains, the MPI standard provides the concept of process topologies. However, the current interface causes problems and requires too much care in its usage: MPI_Dims_create does not take into account the application topology and most implementations of MPI_Cart_create do not consider the underlying network topology and node architecture. To overcome these shortcomings, we defined a new interface that includes application-aware weights to address the communication needs of grid-based applications. The new interface provides a hardware-aware factorization of the processes together with an optimized process mapping onto the underlying hardware resources. The paper describes the underlying implementation, which uses a new multi-level factorization and decomposition approach minimizing slow inter-node communication. Benchmark results show the significant performance gains on multi node NUMA systems.

The effects of scenario-based simulation course training on nurses' communication competence and self-efficacy: a randomized controlled trial.

Leveraging label hierarchy using transfer and multi-task learning: A case study on patent classification

Despite technological advantages in microprocessors and network technology over the last few years, commercially-available networks of workstations (NOW's) contain inherent communication bottlenecks. Traditional layered network protocols will inevitably fail to achieve high throughput if they access data several times. As a result, applications on NOW often fail to observe the performance speed-up that might be expected Network protocols which avoid routing through the kernel can remove this limit on communication performance and support very high transmission speeds, turning NOWs into an attractive alternative to Massively-Parallel Processors. This paper investigates the improvements in performance of an entire NOW system that are possible by changing communication protocols. A commodity workstation cluster with a zero-copy communication protocol (Basic Interface Parallelism) is examined and compared with both conventional protocol implementations and parallel systems in order to understand and assess the improvements in performance achieved.

IA21: Risk prediction models for lung cancer

The aim of this study was to analyse the front-crawl arm-pull kinetics and kinematics, comparing it before and after post-activation potentiation (PAP), and the associations between variables describing of the arm-pull kinetics. Twelve male competitive swimmers were randomly assigned to perform two different warm-ups in a crossover manner: (i) non-PAP (control condition); and (ii) PAP (experimental condition). PAP consisted of 2 × 5 arm-pulls with resistance bands by both upper-limbs. Eight minutes later, participants underwent a 25 m all-out trial in front-crawl arm-pull. Kinetics (i.e., peak thrust, mean thrust and thrust-time integral) and kinematics (i.e., speed and speed fluctuation) were collected by an in-house customised system composed of differential pressure sensors, speedo-meter and underwater camera. There was a significant and large improvement of the arm-pull kinetics after completing the warm-up with PAP sets (0.010 < P < 0.054, 0.50 < d < 0.74). There were non-significant and small effects of PAP on speed (P = 0.307, d = 0.18) and speed fluctuation (P = 0.498, d = 0.04). Correlation coefficients among kinetic variables were significant with large associations (0.51 < R < 0.90, 0.001 < P < 0.088). In conclusion, warm-ups including PAP conditioning sets elicit a large improvement in the thrust, but with small improvement in performance. Variables used to characterise thrust are strongly correlated and hence can be used interchangeably.

This paper presents a study on the concurrent topology optimization of a structure and its material microstructure. A modified optimization model is proposed by introducing microstructure orientation angles as a new type of design variable. The new model is based on the assumptions that a structure is made of a material with the same microstructure, and the material may have a different orientation within the design domain of the structure. The homogenization theory is applied to link the material and structure scales. An additional post-processing technique is developed for modifying the obtained design to avoid local optima caused by the use of orientation angle variables. Numerical examples are presented to illustrate the viability and effectiveness of the proposed model. It is found that significant improvement in structural performance can be achieved by optimizing the orientation of microstructures in concurrent topology optimization of structures and materials. This paper presents a modified concurrent optimization model with three kinds of design variables. At the micro scale, topological design variables are employed for describing the distribution of base material within a material cell so that the material microstructure can be defined. At the macro scale, in addition to topological design variables for describing the distribution of material with microstructure within the design domain, microstructure orientation angles are introduced for defining material axis directions at different locations. Numerical examples have shown that significant improvement in the performance of optimized designs can be achieved.

The goal of this research is to demonstrate that teaching healthcare trainees a formal search or scan pattern for evaluation of the lungs improves their ability to identify pulmonary nodules on chest radiographs (CXRs). A group of physician assistant trainees were randomly assigned to control and experimental groups. Each group was shown two sets of CXRs, each set with a nodule prevalence of approximately 50%. The experimental group received search pattern training between case sets, whereas the control group did not. Both groups were asked to mark nodules when present and indicate their diagnostic confidence. Subject performance at nodule detection was quantified using changes in area under the localization receiver operating characteristic curve ([Formula: see text]). There was no significant improvement in performance between case sets for the control group. There was a significant improvement in subject performance after training for the experimental group, [Formula: see text], [Formula: see text]. These results demonstrate that teaching a search pattern to trainees improves their ability to identify nodules and decreases the number of perceptual errors in nodule identification, and suggest that our knowledge of medical image perception may be used to develop rational tools for the education of healthcare trainees.

This work provides a systematic study of the impact of communication performance on parallel applications in a high performance network of workstations. We develop an experimental system in which the communication latency, overhead, and bandwidth can be independently varied to observe the effects on a wide range of applications. Our results indicate that current efforts to improve cluster communication performance to that of tightly integrated parallel machines results in significantly improved application performance. We show that applications demonstrate strong sensitivity to overhead, slowing down by a factor of 60 on 32 processors when overhead is increased from 3 to 103 &amp;micro;s. Applications in this study are also sensitive to per-message bandwidth, but are surprisingly tolerant of increased latency and lower per-byte bandwidth. Finally, most applications demonstrate a highly linear dependence to both overhead and per-message bandwidth, indicating that further improvements in communication performance will continue to improve application performance.