parameters for i3en.3xlarge (need to add more modes to the benchmark, e.g., time based)

setup_bench_repo_dir.bash

@@ -0,0 +1,39 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+if [ "$(cat /sys/class/block/nvme1n1/device/model)" != "Amazon EC2 NVMe Instance Storage        " ]; then
+    echo "nvme1n1 is not Amazon EC2 NVMe Instance Storage: '$(cat /sys/class/block/nvme1n1/device/model)'"
+    exit 1
+fi
+
+rmdir bench_repo_dir || true
+
+sudo mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0  /dev/nvme1n1
+
+sudo mount /dev/nvme1n1 /mnt
+sudo chown -R "$(id -u)":"$(id -g)" /mnt
+
+mkdir /mnt/bench_repo_dir
+mkdir bench_repo_dir
+sudo mount --bind /mnt/bench_repo_dir bench_repo_dir
+
+mkdir /mnt/test_output
+
+echo run the following commands
+
+cat <<EOF
+    # test suite run
+    export TEST_OUTPUT="/mnt/test_output"
+    DEFAULT_PG_VERSION=15 BUILD_TYPE=release ./scripts/pytest test_runner/performance/test_pageserver.py
+
+    # for interactive use
+    export NEON_REPO_DIR="$(readlink -f ./bench_repo_dir)/repo"
+    cargo build_testing --release
+    ./target/release/neon_local init
+    # ... create tenant, seed it using pgbench
+    # then duplicate the tenant using
+    # poetry run python3 ./test_runner/duplicate_tenant.py TENANT_ID 200 8
+EOF
+
+

test_runner/performance/test_pageserver.py

@@ -42,7 +42,7 @@ def test_getpage_throughput(
     template_tenant, template_timeline = env.neon_cli.create_tenant(conf=tenant_config_cli)
     template_tenant_gen = int(ps_http.tenant_status(template_tenant)["generation"])
     with env.endpoints.create_start("main", tenant_id=template_tenant) as ep:
-        pg_bin.run_capture(["pgbench", "-i", "-s1", ep.connstr()])
+        pg_bin.run_capture(["pgbench", "-i", "-s50", ep.connstr()])
         last_flush_lsn_upload(env, ep, template_tenant, template_timeline)
     ps_http.tenant_detach(template_tenant)
 
@@ -53,7 +53,7 @@ def test_getpage_throughput(
     src_timelines_dir: Path = remote_storage.tenant_path(template_tenant) / "timelines"
     assert src_timelines_dir.is_dir(), f"{src_timelines_dir} is not a directory"
     tenants = [template_tenant]
-    for i in range(0, 10):
+    for i in range(0, 200):
         new_tenant = TenantId.generate()
         tenants.append(new_tenant)
         log.info("Duplicating tenant #%s: %s", i, new_tenant)
@@ -110,7 +110,7 @@ def test_getpage_throughput(
         "--num-tasks",
         "1",
         "--num-requests",
-        "10000",
+        "200000",
         *[str(tenant) for tenant in tenants],
     ]
     log.info(f"command: {' '.join(cmd)}")