Felix Perez

ParFiSys: a parallel file system for MPP

The continuous MPP computing power growth has not been corresponded with an equivalent improvement in I/O bandwidth. As a result, computation, memory and I/O are more unbalanced every year, with more MPP applications being I/O bounded. This paper gives an overview of ParFiSys, a parallel file system developed at the UPM to provide I/O services to the GPMIMD machine, an MPP developed within an ESPRIT project. The main goals of ParFiSys are to provide I/O services to scientific applications requiring high I/O bandwidth, to minimize application porting effort and to exploit the parallelism of generic message-passing multicomputers.

Performance increase mechanisms for parallel and distributed file systems

Abstract This paper presents ParFiSys , a parallel file system developed at the UPM. ParFiSys 1 provides transparent access to several types of distributed file systems, which may be accessed using different mapping functions. Grouped management, parallelization, resource preallocation and write-before-full cache policies, are presented as relevant features of ParFiSys . Finally, some results of ParFiSys evaluation are presented to demonstrate its viability.

High Performance Cache Management for Parallel File Systems

Caching has been intensively used in memory and traditional file systems to improve system performance. However, the use of caching in parallel file systems has been limited to I/O nodes to avoid cache coherence problems. In this paper we present the cache mechanisms implemented in ParFiSys, a parallel file system developed at the UPM. ParFiSys exploits the use of cache, both at processing and I/O nodes, with aggressive pre-fetching and delayed-write techniques. The cache coherence problem is solved by using a dynamic scheme of cache coherence protocols with different sizes and shapes of granularity. Performance results, obtained on an IBM SP2, are presented to demonstrate the advantages offered by the cache management methods used in ParFiSys.

Evaluating ParFiSys : a high-performance parallel and distributed file system

We present an overview of ParFiSys, a coherent parallel file system developed at the UPM to provide I/O services to the GPMIMD machine, an MPP built within the ESPRIT project P-5404. Special emphasis is made on the results obtained during ParFiSys evaluation. They were obtained using several I/O benchmarks (PARKBENCH, IOBENCH, etc.) and several MPP platforms (T800, T9000, etc.) to cover a big spectrum of the ParFiSys features, being specifically oriented to measure throughput for scientific applications I/O patterns. ParFiSys is specially well suited to provide I/O services to scientific applications requiring high I/O bandwidth, to minimize application porting effort, and to exploit the parallelism of generic message-passing multicomputers.

I/O Data Mapping in ParFiSys: Support for High-Performance I/O in Parallel and Distributed Systems

This paper gives an overview of the I/O data mapping mechanisms of ParFiSys. Grouped management and parallelization are presented as relevant features. I/O data mapping mechanisms of ParFiSys, including all levels of the hierarchy, are described in this paper.

A massively parallel and distributed I/O subsystem

This paper presents ParFiSys, a parallel and distributed file system developed at the UPM. It provides several types of distributed file systems, which may be accessed using different mapping functions. Grouped management and parallelization are presented as new features of ParFiSys, not included in other existing parallel and distributed file systems. Moreover, the ParFiSys interface is briefly explained to show the facilities available to general purpose and supercomputing application programmers.

SPICE implementation of nonuniform frequency-dependent transmission lines

A technique for obtaining equivalent circuits of nonuniform single transmission lines, with frequency-dependent parameters is presented. This technique is based on the Branins method of characteristics. Existing commercial CAD software packages, such as SPICE, can be used for this implementation. The starting point of the technique consists of finding the functional relations between the S parameters nonuniform transmission line and the characterization functions of the circuit model proposed which are input impedances -Z/sub 1/(s)- and the propagation functions -F/sub 12/(s), F/sub 21/(s)-. The circuits are synthesized using conventional network synthesis techniques. An examination of the validity of the model is carried out analyzing the distortion of a trapezoidal pulse as it propagates along a microstrip taper line.<<ETX>>

GENOME WIDE ASSOCIATION STUDY FOR HYPERTENSION-INDUCED MICROALBUMINURIA: PP.21.318

Objective: To identify genetic susceptibility variants for hypertension associated microalbuminuria. Methods: The study design was a nested case-control genetic association study. Three hundred and three patients were selected from one prospective cohort recruited in the hypertension clinic of two different hospitals with the objective of identify clinical factors associated with microalbuminuria. All the patients were less than 50y, hypertensive grade I and naïve regarding to the treatment at the time of the recruitment. Cases were those who developed microalbuminuria (UAE ≥30 mg/day) during the follow up and controls were those who remained with an UAE less than 30 mg/day. UAE was assessed in 24 hour urinary and in two or three samples. The Nsp array of the Affymetrix Gene-Chip Human Mapping 500K was used for genotyping, BRLMM algorithm for inferring the genotypes and PLINK software for performing the association analysis after quality control procedures. Results: The final sample size compromised of 201 controls and 102 cases with genotypes for 250000 SNPs. A few polymorphisms and haplotypes resulted in an association with the presence of an increment in UAE expressed as a qualitative or as a quantitative trait. The significantly associated loci were 8p22 for UAE as a qualitative trait and 8q12.1, 2p11.2, 9p22.2 and 18q23 for the quantitative trait. Some possible candidate genes within the significantly associated loci were: MSR1 for the qualitative trait and IMPAD1, REEP1, BNC2, and CNDP1 for the quantitative trait. Figure 1. No caption available. Other group of SNPs did not reach the statistical significance but they seemed to stand out from the rest of SNPs. Within those regions there were also some possible candidate genes such as ELMO, LOX1, NRP1 or PEBP1 among others. Conclusions: GWAS strategy has enabled the identification of some genomic areas with potential influence on UAE in hypertensive patients. Some of them which are mainly involved in lipid metabolism, inflammation or intracellular signaling have been previously reported linked or associated loci with renal damage.

Two receivers for a CPE 64-QAM LMDS

Two prototypes for a Customer Premises Equipment (CPE) for a 64-QAM LMDS in the frequency range of 27.5 - 31.3 GHz have been designed and fabricated. These receivers have been tested in a cellular point-to-multipoint configuration in a urban environment with a radius of 3 Km. The first receiver was designed as a rapid prototype using instrumentation components. The second receiver is a low cost equipment, which is more appropriate for commercial deployment of this system.

Commercial balanced optical front-end for 2.5 Gbit/s continuous phase frequency shift keying (CPFSK) coherent transmission system

In this paper, a balanced optical front-end and IF module for a coherent heterodyne optical receiver are presented. The balanced optical front-end is realized with two commercially available GaInAs/InP p-i-n photodiodes and a full custom designed GaAs MMIC transimpedance amplifier. In this front-end a transimpedance, 3 dB bandwidth, and bandwidth- average input-referred noise current density of 58 dB(Omega) , at 1.7 GHz through 6.3 GHz, and 10 pA(root)Hz, have been measured. The IF module consists of three amplifier stages, a variable attenuator, a pass-band filter, and a broadband coupler. The amplifier stages and the variable attenuator are realized with commercially available MMICs chips. The filter and coupler have been built in hybrid technology. The measured characteristics are 47 dB gain with +/- 1.5 dB ripple, and more than 12 dB input/output port return losses. A transimpedance of 104 dB(Omega) has been measured in this complete optical receiver.