mirror of
https://github.com/neondatabase/neon.git
synced 2026-05-22 07:30:37 +00:00
run test_pageserver_max_throughput_getpage_at_latest_lsn with 1k tenants, compare std-fs with tokio-epoll-uring
This commit is contained in:
Christian Schwarz
36
results.txt
Normal file
36
results.txt
Normal file
@@ -0,0 +1,36 @@
|
||||
run on i3en.3xlarge
|
||||
|
||||
admin@ip-172-31-13-23:[~/neon-main]: du -hs /instance_store/test_output/shared-snapshots/max_throughput_latest_lsn-1000-6/snapshot/local_fs_remote_storage/pageserver/tenants
|
||||
225G /instance_store/test_output/shared-snapshots/max_throughput_latest_lsn-1000-6/snapshot/local_fs_remote_storage/pageserver/tenants
|
||||
|
||||
=> ~2.25x main memory
|
||||
|
||||
admin@ip-172-31-13-23:[~/neon-main]: NEON_ENV_BUILDER_USE_OVERLAYFS_FOR_SNAPSHOTS=1 DEFAULT_PG_VERSION=15 BUILD_TYPE=release ./scripts/pytest test_runner/performance/pageserver/pagebench/test_pageserver_max_throughput_getpage_at_latest_lsn.py
|
||||
|
||||
--------------------------------------------------------------------------------- Benchmark results ---------------------------------------------------------------------------------
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.n_tenants: 1000
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.pgbench_scale: 6
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.duration: 30 s
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config_override.page_cache_size: 134217728 byte
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config_override.max_file_descriptors: 500000
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config.override.virtual_file_io_engine: IoEngine.STD_FS
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.request_count: 2321
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.latency_mean: 8,785.440 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p95: 20,234.239 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99: 20,234.239 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99.9: 20,234.239 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-std-fs].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99.99: 20,234.239 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.n_tenants: 1000
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.pgbench_scale: 6
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.duration: 30 s
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config_override.page_cache_size: 134217728 byte
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config_override.max_file_descriptors: 500000
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.pageserver_config.override.virtual_file_io_engine: IoEngine.TOKIO_EPOLL_URING
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.request_count: 2200
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.latency_mean: 9,046.271 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p95: 16,457.727 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99: 16,457.727 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99.9: 16,457.727 ms
|
||||
test_pageserver_max_throughput_getpage_at_latest_lsn[1000-6-30-tokio-epoll-uring].pageserver_max_throughput_getpage_at_latest_lsn.latency_percentiles.p99.99: 16,457.727 ms
|
||||
|
||||
=========================================================================== 2 passed in 142.33s (0:02:22) ===========================================================================
|
||||
@@ -20,10 +20,10 @@ fi
|
||||
|
||||
# do all the on-disk initialization work now instead of a background kernel thread
|
||||
# so that we're ready for benchmarking right after this line
|
||||
sudo mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 /dev/nvme1n1
|
||||
#sudo mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 /dev/nvme1n1
|
||||
|
||||
MOUNTPOINT=/instance_store
|
||||
sudo mkdir "$MOUNTPOINT"
|
||||
sudo rmdir "$MOUNTPOINT" || sudo mkdir "$MOUNTPOINT"
|
||||
sudo mount /dev/nvme1n1 "$MOUNTPOINT"
|
||||
sudo chown -R "$(id -u)":"$(id -g)" "$MOUNTPOINT"
|
||||
|
||||
@@ -40,7 +40,7 @@ To run your local neon.git build on the instance store volume,
|
||||
run the following commands from the top of the neon.git checkout
|
||||
|
||||
# raise file descriptor limit of your shell and its child processes
|
||||
sudo prlimit -p $$ --nofile=800000:800000
|
||||
sudo prlimit -p \$\$ --nofile=800000:800000
|
||||
|
||||
# test suite run
|
||||
export TEST_OUTPUT="$TEST_OUTPUT"
|
||||
|
||||
@@ -1,7 +1,10 @@
|
||||
import enum
|
||||
import json
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, Tuple
|
||||
|
||||
import toml
|
||||
|
||||
import fixtures.pageserver.many_tenants as many_tenants
|
||||
import pytest
|
||||
from fixtures.benchmark_fixture import MetricReport, NeonBenchmarker
|
||||
@@ -17,6 +20,10 @@ from fixtures.utils import get_scale_for_db, humantime_to_ms
|
||||
from performance.pageserver.util import ensure_pageserver_ready_for_benchmarking
|
||||
|
||||
|
||||
class IoEngine(str, enum.Enum):
|
||||
STD_FS = "std-fs"
|
||||
TOKIO_EPOLL_URING = "tokio-epoll-uring"
|
||||
|
||||
# For reference, the space usage of the snapshots:
|
||||
# admin@ip-172-31-13-23:[~/neon-main]: sudo du -hs /instance_store/test_output/shared-snapshots
|
||||
# 137G /instance_store/test_output/shared-snapshots
|
||||
@@ -27,9 +34,10 @@ from performance.pageserver.util import ensure_pageserver_ready_for_benchmarking
|
||||
# 5.1G /instance_store/test_output/shared-snapshots/max_throughput_latest_lsn-10-6
|
||||
# 76G /instance_store/test_output/shared-snapshots/max_throughput_latest_lsn-100-13
|
||||
# 46G /instance_store/test_output/shared-snapshots/max_throughput_latest_lsn-100-6
|
||||
@pytest.mark.parametrize("ioengine", [IoEngine.STD_FS, IoEngine.TOKIO_EPOLL_URING])
|
||||
@pytest.mark.parametrize("duration", [30])
|
||||
@pytest.mark.parametrize("pgbench_scale", [get_scale_for_db(s) for s in [100, 200]])
|
||||
@pytest.mark.parametrize("n_tenants", [1, 10])
|
||||
@pytest.mark.parametrize("pgbench_scale", [get_scale_for_db(s) for s in [100]])
|
||||
@pytest.mark.parametrize("n_tenants", [1000])
|
||||
@pytest.mark.timeout(
|
||||
10000
|
||||
) # TODO: this value is just "a really high number"; have this per instance type
|
||||
@@ -40,6 +48,7 @@ def test_pageserver_max_throughput_getpage_at_latest_lsn(
|
||||
n_tenants: int,
|
||||
pgbench_scale: int,
|
||||
duration: int,
|
||||
ioengine: IoEngine,
|
||||
):
|
||||
def record(metric, **kwargs):
|
||||
zenbenchmark.record(
|
||||
@@ -60,9 +69,12 @@ def test_pageserver_max_throughput_getpage_at_latest_lsn(
|
||||
# configure cache sizes like in prod
|
||||
page_cache_size = 16384
|
||||
max_file_descriptors = 500000
|
||||
neon_env_builder.pageserver_config_override = (
|
||||
f"page_cache_size={page_cache_size}; max_file_descriptors={max_file_descriptors}"
|
||||
)
|
||||
pageserver_config_override = {
|
||||
"page_cache_size": f"{page_cache_size}",
|
||||
"max_file_descriptors": f"{max_file_descriptors}",
|
||||
"virtual_file_io_engine": f"\"{ioengine}\"",
|
||||
}
|
||||
neon_env_builder.pageserver_config_override = ";".join([f"{k}={v}" for k, v in pageserver_config_override.items()])
|
||||
params.update(
|
||||
{
|
||||
"pageserver_config_override.page_cache_size": (
|
||||
@@ -70,6 +82,7 @@ def test_pageserver_max_throughput_getpage_at_latest_lsn(
|
||||
{"unit": "byte"},
|
||||
),
|
||||
"pageserver_config_override.max_file_descriptors": (max_file_descriptors, {"unit": ""}),
|
||||
"pageserver_config.override.virtual_file_io_engine": (ioengine, {"unit": ""}),
|
||||
}
|
||||
)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user