Rodrigo Virote Kassick

AGIOS: Application-Guided I/O Scheduling for Parallel File Systems

Running data-intensive scientific workflow across multiple data centers faces massive data transfer problem which leads to low efficiency in actual workflow application for scientists. By considering data size and data dependency, we propose a k-means algorithm based initial data placement strategy that places the most related initial data sets into the same data center at workflow preparation stage. During the execution of scientific workflow, by analyzing interdependent relationship between data sets and tasks, we adopt multilevel task replication strategy to reduce volume of intermediate data transfer. The simulation results show that the proposed strategies can effectively reduce data transfer among data centers and improve performance of running data intensive scientific workflows.In remote sensing applications, the traditional procedure for registering a pair of images requires the manual selection of ground control points at significant landmarks of the images. The primary drawback of this approach is that a trained expert is needed to manually select each individual ground control point in the remotely sensed images. This is very laborious and time consuming. In this paper, a new feature-based approach to automate image-to-image registration is proposed for remote sensing applications. This new approach exploits the nonsubsampled contourlet transform to automatically extract a set of control points across spatial and directional resolutions where misalignment between images can be expected to appear. Preliminary experimental results demonstrate the effectiveness and accuracy of the proposed algorithm.For improving the circuit yield, a variety of works based on Dual-Modular Redundancy (DMR) have been studied. Recently, a novel circuit model, Partially Programmable Circuit (PPC), which is made through inserting some Look-Up Tables (LUTs) in conventional logic circuits, has been also proposed. However, they have in common in that they suffer from large area overhead, especially when not all but some faults are specified to be bypassed. Focusing on the fact that an identical copy of parts of the original circuits in DMR and the full programmability of LUTs in PPCs both contain too much functionality for bypassing specified faults, in this paper, we propose three methods to efficiently reduce the functionality of sub-circuits. They can be classified into two approaches, the first one is to reduce the programmability of LUTs in PPCs, and the second one is to apply different ASIC-based sub-circuits, each of which is to bypass a single fault. For the second approach, we propose a novel framework, called Multiple Functional Redundancy (MFR) and its compression method. Experimental results demonstrated that our proposed methods overcome traditional DMR and the original PPC in area overhead while achieving the same yield improvement.

Trace-Based Visualization as a Tool to Understand Applications' I/O Performance in Multi-core Machines

This paper presents the use of trace-based performance visualization of a large scale atmospheric model, the Ocean-Land-Atmosphere Model (OLAM). The trace was obtained with the libRastro library, and the visualization was done with Paj´e. The use of visualization aimed to analyze OLAMs performance and to identify its bottlenecks. Especially, we are interested in the models I/O operations, since it was proved to be the main issue for the models performance. We show that most of the time spent in the output routine is spent in the close operation. With this information, we delayed this operation until the next output phase, obtaining improved I/O performance.

Improving Atmospheric Model Performance on a Multi-Core Cluster System

Numerical models have been used extensively in the last decades to understand and predict weather phenomena and the climate. In general, models are classified according to their operation domain: global (entire Earth) and regional (country, state, etc). Global models have spatial resolution of about 0.2 to 1.5 degrees of latitude and therefore cannot represent very well the scale of regional weather phenomena. Their main limitation is computing power. On the other hand, regional models have higher resolution but are restricted to limited area domains. Forecasting on limited domain demands the knowledge of future atmospheric conditions at domain’s borders. Therefore, regional models require previous execution of global models.

Impact of I/O Coordination on a NFS-Based Parallel File System with Dynamic Reconfiguration

The large gap between processing and I/O speed makes the storage infrastructure of a cluster a great bottleneck for HPC applications. Parallel File Systems propose a solution to this issue by distributing data onto several servers, dividing the load of I/O operations and increasing the available bandwidth. However, most parallel file systems use a fixed number of I/O servers defined during initialization and do not support addition of new resources as applications’ demands grow. With the execution of different applications at the same time, the concurrent access to these resources can impact the performance and aggravate the existing bottleneck. The dNFSp File System proposes a reconfiguration mechanism that aims to include new I/O resources as application’s demands grow. These resources are standard cluster nodes and are dedicated to a single application. This paper presents a study of the I/O performance of this reconfiguration mechanism under two circunstances: the use of several independent processes on a multi-core system or of a single centralized I/O process that coordinates the requests from all instances on a node. We show that the use of coordination can improve performance of applications with regular intervals between I/O phases. For applications with no such intervals, on the other hand, uncoordinated I/O presents better performance.

Atmospheric models hybrid OpenMP/MPI implementation multicore cluster evaluation

Atmospheric models usually demand high processing power and generate large amounts of data. As the degree of parallelism grows, the I/O operations may become the major impacting factor of their performance. This work shows that a hybrid MPI/OpenMP implementation can improve the performance of the atmospheric model ocean-land-atmosphere model (OLAM) on a multicore cluster environment. We show that the hybrid MPI/OpenMP version of OLAM decreases the number of output files, resulting in better performance for I/O operations. We have evaluated OLAM on the parallel file system PVFS and shown that storing the files on PVFS results in lower performance than using the local disks of the cluster nodes due as a consequence of file creation and network concurrency. We have also shown that further parallel optimisations should be included in the hybrid version in order to improve the parallel execution time of OLAM.