Dynamic-CoMPI: dynamic optimization techniques for MPI parallel applications

Abstract

This work presents an optimization of MPI communications, called Dynamic-CoMPI, which uses two techniques in order to reduce the impact of communications and non-contiguous I/O requests in parallel applications. These techniques are independent of the application and complementaries to each other. The first technique is an optimization of the Two-Phase collective I/O technique from ROMIO, called Locality aware strategy for Two-Phase I/O (LA-Two-Phase I/O). In order to increase the locality of the file accesses, LA-Two-Phase I/O employs the Linear Assignment Problem (LAP) for finding an optimal I/O data communication schedule. The main purpose of this technique is the reduction of the number of communications involved in the I/O collective operation. The second technique, calledAdaptive-CoMPI, is based on run-time compression of MPI messages exchanged by applications. Both techniques can be applied on every application, because both of them are transparent for the users. Dynamic-CoMPI has been validated by using several MPI benchmarks and real HPC applications. The results show that, for many of the considered scenarios, important reductions in the execution time are achieved by reducing the size and the number of the messages. Additional benefits of our approach are the reduction of the total communication time and the network contention, thus enhancing, not only performance, but also scalability.

A Quantitative Model for Predicting Cross-application Interference in Virtual Environments

“…After Computing Phase Loop execution, synthetic application performs Communication Phase by executing Communication Phase Loop (lines 10 to 12) whose number of iterations is determined by input parameter β. This phase is based on MP Bench benchmark [16]. From all MPI operations executed by this benchmark, MPI Alltoall was particularly interesting for our work because it is widely used in scientific applications.…”

SANComSim: A Scalable, Adaptive and Non-intrusive Framework to Optimize Performance in Computational Science Applications

“…Dynamic-CoMPI [13] presents an optimization of MPI communications, called Dynamic-CoMPI, which uses different techniques in order to reduce the impact of communications and non-contiguous I/O requests in parallel applications. These techniques are independent of the application and complementary to each other.…”

“…This analysis consists in executing a synthetic benchmark for calculating the time required for compressing/decompressing data using different algorithms. In our previous work [13], we have found that a message has three main properties that affect the performance of compressing/decompressing data: message size; datatype of each message element; and redundancy level of message data.…”

Abstract: Reservation-Based I/O Performance Guarantee for MPI-IO Applications Using Shared Storage Systems

“…Systems overview and the other is run-time adaptive message compression. The benefit of LA-Two-Phase I/O was previously examined with PVFS2 as an underlying storage system [8].…”

“…We are using Dynamic-CoMPI [8] as a MPI-IO implementation and Papio [9] as a shared storage system, which implements parallel I/O and performance reservation. We are developing the ADIO layer to connect these systems and to evaluate the benefits of the reservation-based performance isolation approach.…”