Add interface to generate one more results with one benchmark job

This commit is contained in:
Fabing Li 2024-12-11 09:56:31 +00:00 committed by Tate, Hongliang Tian
parent 5d1c16896a
commit 946e4d5529
5 changed files with 236 additions and 145 deletions

View File

@ -11,7 +11,7 @@ jobs:
runs-on: self-hosted
strategy:
matrix:
benchmark:
benchmarks:
- sysbench/cpu_lat
- sysbench/thread_lat
# Memory-related benchmarks
@ -90,7 +90,7 @@ jobs:
RUSTUP_UPDATE_ROOT: https://mirrors.ustc.edu.cn/rust-static/rustup
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
- name: Set up the environment
run: |
chmod +x test/benchmark/bench_linux_and_aster.sh
@ -99,53 +99,87 @@ jobs:
git config --global http.sslVerify false
git config --global http.version HTTP/1.1
- name: Run benchmark
uses: nick-invision/retry@v2 # Retry the benchmark command in case of failure
- name: Run benchmarks
uses: nick-invision/retry@v3 # Retry the benchmark command in case of failure
with:
timeout_minutes: 20
max_attempts: 3
command: |
make install_osdk
BENCHMARK_TYPE=$(jq -r '.benchmark_type' test/benchmark/${{ matrix.benchmark }}/config.json)
ASTER_SCHEME=$(jq -r '.aster_scheme' test/benchmark/${{ matrix.benchmark }}/config.json)
bash test/benchmark/bench_linux_and_aster.sh ${{ matrix.benchmark }} $BENCHMARK_TYPE $ASTER_SCHEME
bash test/benchmark/bench_linux_and_aster.sh "${{ matrix.benchmarks }}"
BENCHMARK_ARTIFACT=results_$(echo "${{ matrix.benchmarks }}" | tr '/' '-')
echo "BENCHMARK_ARTIFACT=$BENCHMARK_ARTIFACT" >> $GITHUB_ENV
- name: Store benchmark results
uses: actions/upload-artifact@v4
with:
name: ${{ env.BENCHMARK_ARTIFACT }}
path: |
result_*.json
Matrix:
runs-on: ubuntu-latest
needs: Benchmarks # Must run after the Benchmarks job.
if: always() # Always run regardless of whether the previous job was successful or not.
outputs:
benchmarks: ${{ steps.set-matrix.outputs.benchmarks }}
steps:
- name: Download results
uses: actions/download-artifact@v4
with:
pattern: results_*
path: ./results
merge-multiple: true
- name: Set matrix for benchmark results
id: set-matrix
run: |
benchmarks=$(ls results/result_*.json | sed 's/.*result_//' | sed 's/\.json//' | jq -R -s -c 'split("\n")[:-1]')
echo benchmarks=$benchmarks >> $GITHUB_OUTPUT
Results:
runs-on: ubuntu-latest
needs: Matrix
strategy:
matrix:
benchmark: ${{ fromJson(needs.Matrix.outputs.benchmarks) }}
max-parallel: 1
fail-fast: false
steps:
- uses: actions/checkout@v4
- name: Download Benchmark Results
uses: actions/download-artifact@v4
with:
pattern: results_*
path: ./results
merge-multiple: true
- name: Set up benchmark configuration
run: |
ALERT_THRESHOLD=$(jq -r '.alert_threshold' test/benchmark/${{ matrix.benchmark }}/config.json)
ALERT_TOOL=$(jq -r '.alert_tool' test/benchmark/${{ matrix.benchmark }}/config.json)
TITLE=$(jq -r '.title' test/benchmark/${{ matrix.benchmark }}/config.json)
DESCRIPTION=$(jq -r '.description' test/benchmark/${{ matrix.benchmark }}/config.json)
if [ "$ALERT_THRESHOLD" = "null" ]; then
ALERT_THRESHOLD="130%"
fi
if [ "$ALERT_TOOL" = "null" ]; then
ALERT_TOOL="customSmallerIsBetter"
fi
if [ "$TITLE" = "null" ]; then
TITLE="${{ matrix.benchmark }}"
fi
BENCHMARK_DIR=$(echo ${{ matrix.benchmark }} | sed 's/-/\//g')
BENCHMARK_SUITE=$(echo $BENCHMARK_DIR | awk -F'/' '{print $1}')
BENCHMARK_NAME=$(basename "$BENCHMARK_DIR")
BENCH_RESULT="test/benchmark/${BENCHMARK_DIR}/bench_result.json"
[ -f "$BENCH_RESULT" ] || BENCH_RESULT="test/benchmark/${BENCHMARK_DIR}.json"
ALERT_THRESHOLD=$(jq -r '.alert.threshold // "130%"' "$BENCH_RESULT")
ALERT_TOOL=$(jq -r 'if .alert.bigger_is_better == true then "customBiggerIsBetter" else "customSmallerIsBetter" end' "$BENCH_RESULT")
TITLE=$(jq -r '.chart.title // "Undefined"' "$BENCH_RESULT")
DESCRIPTION=$(jq -r '.chart.description // "No description provided"' "$BENCH_RESULT")
echo "BENCHMARK_SUITE=$BENCHMARK_SUITE" >> $GITHUB_ENV
echo "BENCHMARK_NAME=$BENCHMARK_NAME" >> $GITHUB_ENV
echo "ALERT_THRESHOLD=$ALERT_THRESHOLD" >> $GITHUB_ENV
echo "ALERT_TOOL=$ALERT_TOOL" >> $GITHUB_ENV
echo "TITLE=$TITLE" >> $GITHUB_ENV
echo "DESCRIPTION=$DESCRIPTION" >> $GITHUB_ENV
BENCHMARK_NAME=$(basename "${{ matrix.benchmark }}")
echo "BENCHMARK_NAME=$BENCHMARK_NAME" >> $GITHUB_ENV
BENCHMARK_STORE_DIR=$(dirname "${{ matrix.benchmark }}")
echo "BENCHMARK_STORE_DIR=$BENCHMARK_STORE_DIR" >> $GITHUB_ENV
BENCHMARK_SUMMARY=test/benchmark/${BENCHMARK_STORE_DIR}/summary.json
echo "BENCHMARK_SUMMARY=$BENCHMARK_SUMMARY" >> $GITHUB_ENV
- name: Store benchmark results
uses: asterinas/github-action-benchmark@v4
with:
name: ${{ env.BENCHMARK_NAME }}
tool: ${{ env.ALERT_TOOL }}
output-file-path: result_${{ env.BENCHMARK_NAME }}.json
benchmark-data-dir-path: ${{ env.BENCHMARK_STORE_DIR }}
output-file-path: results/result_${{ matrix.benchmark }}.json
benchmark-data-dir-path: ${{ env.BENCHMARK_SUITE }}
github-token: ${{ secrets.BENCHMARK_SECRET }}
gh-repository: 'github.com/asterinas/benchmark'
auto-push: true
@ -154,4 +188,4 @@ jobs:
fail-on-alert: true
chart-title: ${{ env.TITLE }}
chart-description: ${{ env.DESCRIPTION }}
summary-json-path: ${{ env.BENCHMARK_SUMMARY }}
summary-json-path: test/benchmark/${{ env.BENCHMARK_SUITE }}/summary.json

View File

@ -8,61 +8,83 @@ set -o pipefail
# Ensure all dependencies are installed
command -v jq >/dev/null 2>&1 || { echo >&2 "jq is not installed. Aborting."; exit 1; }
# Script directory
BENCHMARK_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
# Set up paths
BENCHMARK_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
source "${BENCHMARK_ROOT}/common/prepare_host.sh"
RESULT_TEMPLATE="${BENCHMARK_ROOT}/result_template.json"
# Source the prepare_host.sh script
source "${BENCHMARK_DIR}/common/prepare_host.sh"
# Parse benchmark results
parse_raw_results() {
local search_pattern="$1"
local result_index="$2"
local result_file="$3"
# Parse the results from the benchmark output
parse_results() {
local benchmark="$1"
local search_pattern="$2"
local result_index="$3"
local linux_output="$4"
local aster_output="$5"
local result_template="$6"
local result_file="$7"
# Extract numeric result from a specific field in the matching line
# Extract and sanitize numeric results
local linux_result aster_result
linux_result=$(awk "/${search_pattern}/ {result=\$$result_index} END {print result}" "${linux_output}" | tr -d '\r' | sed 's/[^0-9.]*//g')
aster_result=$(awk "/${search_pattern}/ {result=\$$result_index} END {print result}" "${aster_output}" | tr -d '\r' | sed 's/[^0-9.]*//g')
linux_result=$(awk "/${search_pattern}/ {result=\$$result_index} END {print result}" "${LINUX_OUTPUT}" | tr -d '\r' | sed 's/[^0-9.]*//g')
aster_result=$(awk "/${search_pattern}/ {result=\$$result_index} END {print result}" "${ASTER_OUTPUT}" | tr -d '\r' | sed 's/[^0-9.]*//g')
# Ensure both results are valid
if [ -z "${linux_result}" ] || [ -z "${aster_result}" ]; then
echo "Error: Failed to parse the results from the benchmark output" >&2
exit 1
fi
echo "Updating the result template with extracted values..."
# Write the results into the template
jq --arg linux_result "${linux_result}" --arg aster_result "${aster_result}" \
'(.[] | select(.extra == "linux_result") | .value) |= $linux_result |
(.[] | select(.extra == "aster_result") | .value) |= $aster_result' \
"${result_template}" > "${result_file}"
"${RESULT_TEMPLATE}" > "${result_file}"
}
# Run the benchmark on Linux and Asterinas
# Generate a new result template based on unit and legend
generate_template() {
local unit="$1"
local legend="$2"
# Replace placeholders with actual system names
local linux_legend=${legend//"{system}"/"Linux"}
local asterinas_legend=${legend//"{system}"/"Asterinas"}
# Generate the result template JSON
jq -n --arg linux "$linux_legend" --arg aster "$asterinas_legend" --arg unit "$unit" '[
{ "name": $linux, "unit": $unit, "value": 0, "extra": "linux_result" },
{ "name": $aster, "unit": $unit, "value": 0, "extra": "aster_result" }
]' > "${RESULT_TEMPLATE}"
}
# Extract the result file path based on benchmark location
extract_result_file() {
local bench_result="$1"
local relative_path="${bench_result#*/benchmark/}"
local first_dir="${relative_path%%/*}"
local filename=$(basename "$bench_result")
# Handle different naming conventions for result files
if [[ "$filename" == bench_* ]]; then
local second_part=$(dirname "$bench_result" | awk -F"/benchmark/$first_dir/" '{print $2}' | cut -d'/' -f1)
echo "result_${first_dir}-${second_part}.json"
else
echo "result_${relative_path//\//-}"
fi
}
# Run the specified benchmark with optional scheme
run_benchmark() {
local benchmark="$1"
local benchmark_type="$2"
local run_mode="$2"
local aster_scheme="$3"
local search_pattern="$4"
local result_index="$5"
local linux_output="${BENCHMARK_DIR}/linux_output.txt"
local aster_output="${BENCHMARK_DIR}/aster_output.txt"
local result_template="${BENCHMARK_DIR}/${benchmark}/result_template.json"
local benchmark_name=$(basename "${benchmark}")
local benchmark_root=$(dirname "${benchmark}")
local result_file="result_${benchmark_name}.json"
echo "Preparing libraries..."
prepare_libs
# Set up Asterinas scheme if specified
local aster_scheme_cmd=""
if [ -n "$aster_scheme" ] && [ "$aster_scheme" != "null" ]; then
aster_scheme_cmd="SCHEME=${aster_scheme}"
fi
# Prepare commands for Asterinas and Linux
local asterinas_cmd="make run BENCHMARK=${benchmark} ${aster_scheme_cmd} ENABLE_KVM=1 RELEASE_LTO=1 NETDEV=tap VHOST=on 2>&1"
local linux_cmd="/usr/local/qemu/bin/qemu-system-x86_64 \
--no-reboot \
@ -72,65 +94,97 @@ run_benchmark() {
-cpu Icelake-Server,-pcid,+x2apic \
--enable-kvm \
-kernel ${LINUX_KERNEL} \
-initrd ${BENCHMARK_DIR}/../build/initramfs.cpio.gz \
-drive if=none,format=raw,id=x0,file=${BENCHMARK_DIR}/../build/ext2.img \
-initrd ${BENCHMARK_ROOT}/../build/initramfs.cpio.gz \
-drive if=none,format=raw,id=x0,file=${BENCHMARK_ROOT}/../build/ext2.img \
-device virtio-blk-pci,bus=pcie.0,addr=0x6,drive=x0,serial=vext2,disable-legacy=on,disable-modern=off,queue-size=64,num-queues=1,request-merging=off,backend_defaults=off,discard=off,write-zeroes=off,event_idx=off,indirect_desc=off,queue_reset=off \
-append 'console=ttyS0 rdinit=/benchmark/common/bench_runner.sh ${benchmark} linux mitigations=off hugepages=0 transparent_hugepage=never quiet' \
-netdev tap,id=net01,script=${BENCHMARK_DIR}/../../tools/net/qemu-ifup.sh,downscript=${BENCHMARK_DIR}/../../tools/net/qemu-ifdown.sh,vhost=on \
-netdev tap,id=net01,script=${BENCHMARK_ROOT}/../../tools/net/qemu-ifup.sh,downscript=${BENCHMARK_ROOT}/../../tools/net/qemu-ifdown.sh,vhost=on \
-device virtio-net-pci,netdev=net01,disable-legacy=on,disable-modern=off,csum=off,guest_csum=off,ctrl_guest_offloads=off,guest_tso4=off,guest_tso6=off,guest_ecn=off,guest_ufo=off,host_tso4=off,host_tso6=off,host_ecn=off,host_ufo=off,mrg_rxbuf=off,ctrl_vq=off,ctrl_rx=off,ctrl_vlan=off,ctrl_rx_extra=off,guest_announce=off,ctrl_mac_addr=off,host_ufo=off,guest_uso4=off,guest_uso6=off,host_uso=off \
-nographic \
2>&1"
case "${benchmark_type}" in
# Run the benchmark depending on the mode
case "${run_mode}" in
"guest_only")
echo "Running benchmark ${benchmark} on Asterinas..."
eval "$asterinas_cmd" | tee ${aster_output}
eval "$asterinas_cmd" | tee ${ASTER_OUTPUT}
prepare_fs
echo "Running benchmark ${benchmark} on Linux..."
eval "$linux_cmd" | tee ${linux_output}
eval "$linux_cmd" | tee ${LINUX_OUTPUT}
;;
"host_guest")
echo "Running benchmark ${benchmark} on host and guest..."
bash "${BENCHMARK_DIR}/common/host_guest_bench_runner.sh" \
"${BENCHMARK_DIR}/${benchmark}" \
bash "${BENCHMARK_ROOT}/common/host_guest_bench_runner.sh" \
"${BENCHMARK_ROOT}/${benchmark}" \
"${asterinas_cmd}" \
"${linux_cmd}" \
"${aster_output}" \
"${linux_output}"
"${ASTER_OUTPUT}" \
"${LINUX_OUTPUT}"
;;
*)
echo "Error: Unknown benchmark type '${benchmark_type}'" >&2
echo "Error: Unknown benchmark type '${run_mode}'" >&2
exit 1
;;
esac
echo "Parsing results..."
parse_results "$benchmark" "$search_pattern" "$result_index" "$linux_output" "$aster_output" "$result_template" "$result_file"
echo "Cleaning up..."
rm -f "${linux_output}"
rm -f "${aster_output}"
}
# Main
# Parse the benchmark configuration
parse_results() {
local bench_result="$1"
BENCHMARK="$1"
if [ -z "$2" ] || [ "$2" = "null" ]; then
BENCHMARK_TYPE="guest_only"
else
BENCHMARK_TYPE="$2"
fi
ASTER_SCHEME="$3"
local search_pattern=$(jq -r '.result_extraction.search_pattern // empty' "$bench_result")
local result_index=$(jq -r '.result_extraction.result_index // empty' "$bench_result")
local unit=$(jq -r '.chart.unit // empty' "$bench_result")
local legend=$(jq -r '.chart.legend // {system}' "$bench_result")
echo "Running benchmark ${BENCHMARK}..."
pwd
if [ ! -d "$BENCHMARK_DIR/$BENCHMARK" ]; then
echo "Error: Benchmark directory not found" >&2
generate_template "$unit" "$legend"
parse_raw_results "$search_pattern" "$result_index" "$(extract_result_file "$bench_result")"
}
# Clean up temporary files
cleanup() {
echo "Cleaning up..."
rm -f "${LINUX_OUTPUT}" "${ASTER_OUTPUT}" "${RESULT_TEMPLATE}"
}
# Main function to coordinate the benchmark run
main() {
local benchmark="$1"
if [[ -z "${BENCHMARK_ROOT}/${benchmark}" ]]; then
echo "Error: No benchmark specified" >&2
exit 1
fi
fi
echo "Running benchmark $benchmark..."
search_pattern=$(jq -r '.search_pattern' "$BENCHMARK_DIR/$BENCHMARK/config.json")
result_index=$(jq -r '.result_index' "$BENCHMARK_DIR/$BENCHMARK/config.json")
# Determine the run mode (host-only or host-guest)
local run_mode="guest_only"
[[ -f "${BENCHMARK_ROOT}/${benchmark}/host.sh" ]] && run_mode="host_guest"
run_benchmark "$BENCHMARK" "$BENCHMARK_TYPE" "$ASTER_SCHEME" "$search_pattern" "$result_index"
local bench_result="${BENCHMARK_ROOT}/${benchmark}/bench_result.json"
local aster_scheme
if [[ -f "$bench_result" ]]; then
aster_scheme=$(jq -r '.runtime_config.aster_scheme // ""' "$bench_result")
else
for job in "${BENCHMARK_ROOT}/${benchmark}"/bench_results/*; do
[[ -f "$job" ]] && aster_scheme=$(jq -r '.runtime_config.aster_scheme // ""' "$job") && break
done
fi
echo "Benchmark completed successfully."
# Run the benchmark
run_benchmark "$benchmark" "$run_mode" "$aster_scheme"
# Parse results if benchmark configuration exists
if [[ -f "$bench_result" ]]; then
parse_results "$bench_result"
else
for job in "${BENCHMARK_ROOT}/${benchmark}"/bench_results/*; do
[[ -f "$job" ]] && parse_results "$job"
done
fi
# Cleanup temporary files
cleanup
echo "Benchmark completed successfully."
}
main "$@"

View File

@ -5,12 +5,12 @@
set -e
BENCHMARK_DIR="/benchmark"
BENCHMARK_ROOT="/benchmark"
READY_MESSAGE="The VM is ready for the benchmark."
BENCH_NAME=$1
BENCHMARK_NAME=$1
SYSTEM="${2:-asterinas}"
echo "Running benchmark: ${BENCH_NAME} on ${SYSTEM}"
echo "Running benchmark: ${BENCHMARK_NAME} on ${SYSTEM}"
print_help() {
echo "Usage: $0 <benchmark_name> <system_type>"
@ -19,17 +19,17 @@ print_help() {
}
# Validate arguments
check_bench_name() {
if [ -z "${BENCH_NAME}" ] || [ -z "${SYSTEM}" ]; then
check_benchmark_name() {
if [ -z "${BENCHMARK_NAME}" ] || [ -z "${SYSTEM}" ]; then
echo "Error: Invalid arguments."
print_help
exit 1
fi
local full_path="${BENCHMARK_DIR}/${BENCH_NAME}"
local full_path="${BENCHMARK_ROOT}/${BENCHMARK_NAME}"
if ! [ -d "${full_path}" ]; then
echo "Directory '${BENCH_NAME}' does not exist in the benchmark directory."
echo "Directory '${BENCHMARK_NAME}' does not exist in the benchmark directory."
print_help
exit 1
fi
@ -63,7 +63,7 @@ prepare_system() {
main() {
# Check if the benchmark name is valid
check_bench_name
check_benchmark_name
# Prepare the system
prepare_system
@ -73,9 +73,9 @@ main() {
echo "${READY_MESSAGE}"
# Run the benchmark
BENCH_SCRIPT=${BENCHMARK_DIR}/${BENCH_NAME}/run.sh
chmod +x ${BENCH_SCRIPT}
${BENCH_SCRIPT}
BENCHMARK_SCRIPT=${BENCHMARK_ROOT}/${BENCHMARK_NAME}/run.sh
chmod +x ${BENCHMARK_SCRIPT}
${BENCHMARK_SCRIPT}
# Shutdown explicitly if running on Linux
if [ "$SYSTEM" = "linux" ]; then

View File

@ -14,8 +14,8 @@ LINUX_OUTPUT=$5
READY_MESSAGE="The VM is ready for the benchmark."
# Import the common functions
BENCHMARK_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)/../"
source "${BENCHMARK_DIR}/common/prepare_host.sh"
BENCHMARK_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)/../"
source "${BENCHMARK_ROOT}/common/prepare_host.sh"
if [[ "$BENCHMARK_PATH" =~ "iperf" ]]; then
# Persist Iperf port

View File

@ -5,14 +5,17 @@
set -e
set -o pipefail
# Set BENCHMARK_DIR to the parent directory of the current directory if it is not set
BENCHMARK_DIR="${BENCHMARK_DIR:-$(cd "$(dirname "${BASH_SOURCE[0]}")/.." &>/dev/null && pwd)}"
# Set BENCHMARK_ROOT to the parent directory of the current directory if it is not set
BENCHMARK_ROOT="${BENCHMARK_ROOT:-$(cd "$(dirname "${BASH_SOURCE[0]}")/.." &>/dev/null && pwd)}"
# Set the log file
LINUX_OUTPUT="${BENCHMARK_ROOT}/linux_output.txt"
ASTER_OUTPUT="${BENCHMARK_ROOT}/aster_output.txt"
# Dependencies for Linux
LINUX_DEPENDENCIES_DIR="/opt/linux_binary_cache"
LINUX_KERNEL="${LINUX_DEPENDENCIES_DIR}/vmlinuz"
LINUX_KERNEL_VERSION="5.15.0-105"
LINUX_MODULES_DIR="${BENCHMARK_DIR}/../build/initramfs/lib/modules/${LINUX_KERNEL_VERSION}/kernel"
WGET_SCRIPT="${BENCHMARK_DIR}/../../tools/atomic_wget.sh"
LINUX_MODULES_DIR="${BENCHMARK_ROOT}/../build/initramfs/lib/modules/${LINUX_KERNEL_VERSION}/kernel"
WGET_SCRIPT="${BENCHMARK_ROOT}/../../tools/atomic_wget.sh"
# Prepare Linux kernel and modules
prepare_libs() {
@ -39,6 +42,6 @@ prepare_libs() {
# Prepare fs for Linux
prepare_fs() {
# Disable unsupported ext2 features of Asterinas on Linux to ensure fairness
mke2fs -F -O ^ext_attr -O ^resize_inode -O ^dir_index ${BENCHMARK_DIR}/../build/ext2.img
mke2fs -F -O ^ext_attr -O ^resize_inode -O ^dir_index ${BENCHMARK_ROOT}/../build/ext2.img
make initramfs
}