fio_bench.sh

#!/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)

# Ensure target directory exists
mkdir -p "$DIR"

echo "========================================================================"
echo " 🏎️  STORAGE PERFORMANCE AUDIT"
echo " 🖥️  Node: $(hostname)"
echo " ⚙️  Config: $JOBS threads | ${FILE_SIZE} per thread | ${RUNTIME} per test"
echo "========================================================================"

run_fio() {
  local label=$1 rw=$2 bs=$3
  printf "  %-15s ... " "$label"
  
  # --direct=1 bypasses local RAM cache to test actual storage speed
  # --fallocate=none prevents slow pre-zeroing on thin-provisioned storage
  result=$(fio --name=bench --group_reporting --time_based=1 \
    --runtime=$RUNTIME --ramp_time=$RAMP \
    --size=$FILE_SIZE --numjobs=$JOBS --ioengine=libaio --direct=1 \
    --iodepth=32 --rw=$rw --bs=$bs --directory=$DIR \
    --thread --fallocate=none --output-format=json 2>/dev/null)

  if [ $? -ne 0 ]; then echo "FAILED"; return; fi

  # Parse JSON results (Units: BW in Bytes, Latency in Nanoseconds)
  stats=$(echo "$result" | python3 -c "
import sys, json
try:
    j = json.load(sys.stdin)['jobs'][0]
    res = j['read'] if j['read']['bw_bytes'] > 0 else j['write']
    print(f\"{res['bw_bytes']} {res['iops']} {res['clat_ns']['percentile']['99.000000']}\")
except: print('0 0 0')")

  read bw iops p99 <<< "$stats"
  
  bw_gb=$(python3 -c "print(f'{$bw / 1e9:.2f}')")
  bw_mb=$(python3 -c "print(f'{$bw / 1e6:.0f}')")
  iops_k=$(python3 -c "print(f'{$iops / 1e3:.1f}')")
  p99_ms=$(python3 -c "print(f'{$p99 / 1e6:.1f}')")

  if [ "$bs" == "1m" ]; then
    printf "✅ %6s GB/s | %8s K IOPS | p99 Latency: %s ms\n" "$bw_gb" "$iops_k" "$p99_ms"
  else
    printf "✅ %6s MB/s | %8s K IOPS | p99 Latency: %s ms\n" "$bw_mb" "$iops_k" "$p99_ms"
  fi
}

# Execute tests in logical order (Write first to initialize data blocks)
run_fio "Seq Write 1M"  write 1m
run_fio "Seq Read  1M"  read  1m
run_fio "Rand Write 4K" randwrite 4k
run_fio "Rand Read  4K" randread  4k

# Cleanup temporary test files
rm -f "$DIR"/bench.* 2>/dev/null
echo "========================================================================"
echo " Audit Complete."