aahmed-se

#!/bin/bash
# 🏎️ CLUSTER STORAGE AUDIT (Sequential & Random Suite)
# Usage: ./storage_bench.sh /path/to/mount

DIR=${1:-/data_vast/bench_$(hostname)}
JOBS=16          # Number of simultaneous threads
RUNTIME=10s       # Duration of each of the 4 tests
RAMP=5s          # Warm-up time to saturate controllers
FILE_SIZE=4G     # Data per thread (Total per node = 64GB)

aahmed-se

#!/bin/bash
# 🏎️ CLUSTER STORAGE AUDIT (Sequential & Random Suite)
# Usage: ./storage_bench.sh /path/to/mount

DIR=${1:-/data_vast/bench_$(hostname)}
JOBS=16          # Number of simultaneous threads
RUNTIME=10s       # Duration of each of the 4 tests
RAMP=5s          # Warm-up time to saturate controllers
FILE_SIZE=4G     # Data per thread (Total per node = 64GB)

aahmed-se

#!/bin/bash
# =============================================================================
# 🏎️  NFS STORAGE BENCHMARK — Audio File I/O Profile
# =============================================================================
#
# PURPOSE:
#   Benchmarks an NFS (or similar networked) filesystem for a distributed ML
#   training pipeline that consumes large volumes of WAV/FLAC audio files.
#
# WHAT THIS TESTS:

aahmed-se

#!/bin/bash
# =============================================================================
# 🏎️  NFS STORAGE BENCHMARK — Audio File I/O Profile (PyTorch-Realistic)
# =============================================================================
#
# PURPOSE:
#   Benchmarks an NFS (or similar networked) filesystem for ML training pipelines
#   that consume large numbers of small WAV/FLAC audio files using PyTorch-style
#   dataloaders.
#

aahmed-se

#!/bin/bash
# =============================================================================
# 🏎️  NFS STORAGE BENCHMARK — Audio File I/O Profile (PyTorch-Realistic)
# =============================================================================
#
# PURPOSE:
#   Benchmarks an NFS (or similar networked) filesystem for ML training pipelines
#   that consume large numbers of small WAV/FLAC audio files using PyTorch-style
#   dataloaders.
#