-
Notifications
You must be signed in to change notification settings - Fork 561
add agentic benchmarking on gke #6772
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Open
george-kalisse-sada
wants to merge
2
commits into
GoogleCloudPlatform:master
Choose a base branch
from
george-kalisse-sada:sada-gke-agentic-benchmarking
base: master
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit
Hold shift + click to select a range
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -12,3 +12,5 @@ | |
| /.idea | ||
| /*git_ignore* | ||
| .DS_Store | ||
| .adk | ||
| tmp/ | ||
171 changes: 171 additions & 0 deletions
171
perfkitbenchmarker/data/k8s_agents/config/gke-benchmark.conf
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,171 @@ | ||
| #!/bin/bash | ||
| # | ||
| # Agentic Workload Benchmarking configuration file for GKE | ||
| # Adapted from nginx DPv2 baseline for Python Sandbox & Chromium Simulation | ||
| # | ||
| # Override machine type and cluster suffix via environment variables: | ||
| # MACHINE_TYPE=c4d-standard-8 CLUSTER_SUFFIX=c4d bash setup_infrastructure_gke.sh | ||
| # | ||
| # Supported profiles: | ||
| # MACHINE_TYPE=c3-standard-192-metal CLUSTER_SUFFIX=c3metal | ||
| # MACHINE_TYPE=c4-standard-8 CLUSTER_SUFFIX=c4 (default) | ||
| # MACHINE_TYPE=c4d-standard-8 CLUSTER_SUFFIX=c4d | ||
| # MACHINE_TYPE=c4a-standard-8 CLUSTER_SUFFIX=c4a (ARM64) | ||
|
|
||
| USER_NAME_PREFIX=${USER%%.*} | ||
|
|
||
| # GCP Project (MUST be set before running any script) | ||
| PROJECT_ID="your-project-id" | ||
| REGION="us-central1" | ||
| ZONE="us-central1-a" | ||
|
|
||
| # Google/ADK aliases (derived from canonical names above) | ||
| # These are used by envsubst for the K8s manifest and by the ADK agent. | ||
| GOOGLE_CLOUD_PROJECT="${PROJECT_ID}" | ||
| GOOGLE_CLOUD_LOCATION="${REGION}" | ||
|
|
||
| # Network Configuration | ||
| VPC_NAME="${USER_NAME_PREFIX}-agentic-vpc" | ||
| SUBNET_NAME="${USER_NAME_PREFIX}-agentic-subnet" | ||
| SUBNET_CIDR="10.134.20.0/24" | ||
| LAPTOP_IP="$(curl -s ifconfig.me)/32" # PUBLIC IP to access the target (dynamically detected) | ||
| # Cloud Router and NAT Configuration | ||
| ROUTER_NAME="${USER_NAME_PREFIX}-agentic-nat-router" | ||
| NAT_NAME="${USER_NAME_PREFIX}-agentic-nat-config" | ||
|
|
||
| # GKE Cluster Configuration | ||
| CLUSTER_SUFFIX="${CLUSTER_SUFFIX:-c4}" | ||
| CLUSTER_NAME="${USER_NAME_PREFIX}-agentic-${CLUSTER_SUFFIX}" | ||
| GKE_VERSION="1.35.3-gke.1389000" | ||
| USE_CONNECT_GATEWAY="${USE_CONNECT_GATEWAY:-true}" # Use Connect Gateway for kubectl access | ||
| # Set to "false" to use direct public endpoint | ||
|
|
||
| # ========================================================================= | ||
| # Machine Type Configuration (overridable via MACHINE_TYPE env var) | ||
| # ========================================================================= | ||
| MACHINE_TYPE="${MACHINE_TYPE:-c4-standard-8}" | ||
|
|
||
| # Derive disk type from machine family: | ||
| # C3 → pd-balanced, C4/C4D/C4A → hyperdisk-balanced | ||
| _MACHINE_FAMILY="${MACHINE_TYPE%%-*}" # e.g. "c4" from "c4-standard-8" | ||
| case "${_MACHINE_FAMILY}" in | ||
| c3) _DISK_TYPE="pd-balanced" ;; | ||
| *) _DISK_TYPE="hyperdisk-balanced" ;; | ||
| esac | ||
|
|
||
| # Derive target architecture from machine family: | ||
| # C4A → arm64, everything else → amd64 | ||
| case "${_MACHINE_FAMILY}" in | ||
| c4a) _TARGET_ARCH="arm64" ;; | ||
| *) _TARGET_ARCH="amd64" ;; | ||
| esac | ||
|
|
||
| # Derive unique master CIDR per cluster (each private cluster needs its own /28): | ||
| # c4 → 172.16.0.0/28, c4d → 172.16.0.16/28, c4a → 172.16.0.32/28, c3metal → 172.16.0.48/28 | ||
| case "${CLUSTER_SUFFIX}" in | ||
| c4) MASTER_IPV4_CIDR="172.16.0.0/28" ;; | ||
| c4d) MASTER_IPV4_CIDR="172.16.0.16/28" ;; | ||
| c4a) MASTER_IPV4_CIDR="172.16.0.32/28" ;; | ||
| c3metal) MASTER_IPV4_CIDR="172.16.0.48/28" ;; | ||
| *) MASTER_IPV4_CIDR="172.16.0.64/28" ;; # fallback for future clusters | ||
| esac | ||
|
|
||
| DEFAULT_POOL_MACHINE_TYPE="${MACHINE_TYPE}" | ||
| DEFAULT_POOL_DISK_TYPE="${_DISK_TYPE}" | ||
| DEFAULT_POOL_DISK_SIZE="50" # Disk size in GB | ||
| DEFAULT_POOL_NODE_COUNT="1" # Number of nodes in the default pool | ||
|
|
||
| # ========================================================================= | ||
| # Agentic Workload NodePools | ||
| # ========================================================================= | ||
|
|
||
| # Sandbox NodePool (Python + Chromium workloads with gVisor) | ||
| SANDBOX_NODE_POOL_NAME="agentic-sandbox-pool" | ||
| SANDBOX_MACHINE_TYPE="${MACHINE_TYPE}" # Same as default pool (overridable) | ||
| SANDBOX_DISK_SIZE="100" | ||
| SANDBOX_DISK_TYPE="${_DISK_TYPE}" # Derived from machine family | ||
| SANDBOX_NODE_COUNT="1" | ||
| SANDBOX_MAX_PODS_PER_NODE="250" # Raise from default 110 to avoid GKE pod limit as density ceiling | ||
| SANDBOX_ENABLE_GVISOR="true" # Enable GKE Sandbox (gVisor) on this pool | ||
|
|
||
| AGENT_SANDBOX_VERSION="v0.4.6" | ||
|
|
||
| # ========================================================================= | ||
| # Workload Configuration | ||
| # ========================================================================= | ||
| AGENTIC_NAMESPACE="agentic" | ||
|
|
||
| # Python Sandbox Workload | ||
| PYTHON_IMAGE="python:3.11-slim" | ||
| PYTHON_POD_NAME="python-sandbox" | ||
| PYTHON_REPLICAS="1" # Start with 1; sweep for density tests | ||
| PYTHON_CPU_REQUEST="1" | ||
| PYTHON_CPU_LIMIT="2" | ||
| PYTHON_MEMORY_REQUEST="1Gi" | ||
| PYTHON_MEMORY_LIMIT="4Gi" | ||
|
|
||
| # Chromium Browser Simulation Workload | ||
| CHROMIUM_IMAGE="${REGION}-docker.pkg.dev/${PROJECT_ID}/agent-sandbox/chrome-sandbox:${_TARGET_ARCH}" | ||
| CHROMIUM_POD_NAME="chromium-sandbox" | ||
| CHROMIUM_REPLICAS="1" # Start with 1; sweep for density tests | ||
|
|
||
| # Mock LLM Coordinator | ||
| MOCK_LLM_IMAGE="python:3.11-slim" | ||
| MOCK_LLM_POD_NAME="mock-llm-coordinator" | ||
| MOCK_LLM_PORT="8080" | ||
|
|
||
| # ========================================================================= | ||
| # Benchmark Parameters | ||
| # ========================================================================= | ||
|
|
||
| # Python Density Benchmark (UC-B) | ||
| SAMPLE_COUNT="20" # Samples per sandbox session | ||
| SAMPLE_WARMUP="0" # Warmup samples (excluded from stats) | ||
|
|
||
| # Payload Transfer Benchmark (UC-D) | ||
| PAYLOAD_SIZE_MB="1" # Default payload size in MB | ||
| PAYLOAD_ITERATIONS="20" # Transfer iterations per session | ||
|
|
||
| # Chromium Benchmark | ||
| CHROMIUM_TASK_COUNT="10" # Number of browser tasks per run | ||
| CHROMIUM_WARMUP_TASKS="2" | ||
|
|
||
| # General | ||
| BENCHMARK_DURATION="300" # Duration in seconds per test | ||
| NOTE="agentic-V0-gVisor-DPv2-baseline" | ||
|
|
||
| # ========================================================================= | ||
| # Logging | ||
| # ========================================================================= | ||
| # Log directory — defaults to tmp/ inside the repo (gitignored). | ||
| # Override by setting BASE_LOG_DIR before sourcing this file, | ||
| # e.g. export BASE_LOG_DIR="$HOME/agentic-logs" to keep logs outside the repo. | ||
| _REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../../../.." && pwd)" | ||
| BASE_LOG_DIR="${BASE_LOG_DIR:-${_REPO_ROOT}/tmp/agentic-logs}" | ||
| WRAPPER_LOG_DIR="${BASE_LOG_DIR}/wrapper_logs" | ||
|
|
||
| LOG_PATH="logs" | ||
| LOG_LEVEL="info" | ||
|
|
||
| # ========================================================================= | ||
| # ADK Agent Deployment | ||
| # ========================================================================= | ||
| ADK_REPO_NAME="adk-repo" # Artifact Registry repository name | ||
| ADK_IMAGE_NAME="adk-agent" # Container image name | ||
| GOOGLE_GENAI_USE_VERTEXAI="true" | ||
| ADK_IMAGE_PATH="${REGION}-docker.pkg.dev/${PROJECT_ID}/${ADK_REPO_NAME}/${ADK_IMAGE_NAME}:${_TARGET_ARCH}" | ||
| ADK_K8S_SA="adk-agent-sa" # Kubernetes service account for the agent | ||
| CLOUD_BUILD_SA="adk-cloud-build-sa" # Service account for Cloud Build submissions | ||
|
|
||
| # Sandbox Router & Warm Pool | ||
| SANDBOX_ROUTER_IMAGE="${REGION}-docker.pkg.dev/${PROJECT_ID}/agent-sandbox/sandbox-router:${_TARGET_ARCH}" | ||
| WARMPOOL_REPLICAS="2" # Number of pre-warmed sandbox pods | ||
|
|
||
| # ========================================================================= | ||
| # Pod Snapshot Configuration (UC-A: Cold Start & Snapshot Pressure Test) | ||
| # ========================================================================= | ||
| ENABLE_POD_SNAPSHOTS="true" # Enable pod snapshots feature on cluster | ||
| SNAPSHOTS_BUCKET_NAME="agent-sandbox-snapshots-${PROJECT_ID}" | ||
| SNAPSHOT_KSA_NAME="pod-snapshot-sa" # KSA for snapshot storage access | ||
| SNAPSHOT_FOLDER="benchmark-snapshots" # Managed folder inside the bucket | ||
| SNAPSHOT_PRELOAD_MB="10" # Default memory preload for snapshot sizing |
70 changes: 70 additions & 0 deletions
70
perfkitbenchmarker/data/k8s_agents/config/native_provision_config.yaml
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,70 @@ | ||
| # Native PKB Provision Config for Agentic Benchmarks | ||
| # Used with --gke_provision_mode=native | ||
| # | ||
| # Prerequisites (run once before PKB): | ||
| # python tools/agentic-benchmark/scripts/prerequisite_setup.py \ | ||
| # --project_id=<project> --machine_type=<machine> | ||
| # | ||
| # IMPORTANT: Do NOT pass --gce_subnet_name on the command line. | ||
| # PKB incorrectly resolves it as the --network value. Instead, pass the | ||
| # subnet via --gke_additional_flags on the command line. | ||
| # | ||
| # Usage (provision): | ||
| # python pkb.py --benchmarks=gke_python_density \ | ||
| # --gke_provision_mode=native \ | ||
| # --benchmark_config_file=k8s_agents/config/native_provision_config.yaml \ | ||
| # --gce_network_name=<user>-agentic-vpc \ | ||
| # --gce_subnet_region=us-central1 \ | ||
| # --zone=us-central1-a \ | ||
| # --project=<project> \ | ||
| # --owner=<owner> \ | ||
| # --container_cluster_version=1.35.3-gke.1389000 \ | ||
| # --gke_additional_flags="--subnetwork=<user>-agentic-subnet,--workload-pool=<project>.svc.id.goog" | ||
| # | ||
| # For sweeps (cluster pre-exists, PKB skips provision/teardown): | ||
| # The sweep bridge injects --run_stage=run,cleanup automatically. | ||
|
|
||
| gke_python_density: | ||
|
Collaborator
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Internally we put a lot of this info but externally it is useful.. it's probably a good addition. |
||
| flags: | ||
| # Force gcloud beta for preview features (pod snapshots) | ||
| gke_use_beta: true | ||
|
|
||
| # Cluster-level additional flags (appended to gcloud [beta] container clusters create) | ||
| # NOTE: --subnetwork and --workload-pool are user/project-specific. | ||
| # Pass them on the command line via --gke_additional_flags=... (comma-separated). | ||
| gke_additional_flags: | ||
| - "--enable-pod-snapshots" | ||
| - "--enable-dataplane-v2" | ||
| - "--enable-private-nodes" | ||
| - "--enable-ip-alias" | ||
| - "--master-ipv4-cidr=172.16.0.0/28" | ||
|
|
||
| # Node-pool-level additional flags (appended to gcloud container node-pools create) | ||
| gke_additional_nodepool_flags: | ||
| - "--max-pods-per-node=250" | ||
|
|
||
| # Standard PKB GKE flags | ||
| container_cluster_version: "1.35.3-gke.1389000" | ||
| gke_enable_shielded_nodes: false | ||
|
|
||
| container_cluster: | ||
| cloud: GCP | ||
| type: Kubernetes | ||
| vm_count: 1 | ||
| vm_spec: | ||
| GCP: | ||
| machine_type: c4-standard-8 | ||
| zone: us-central1-a | ||
| boot_disk_type: hyperdisk-balanced | ||
| boot_disk_size: 50 | ||
| nodepools: | ||
| sandbox: | ||
| vm_count: 1 | ||
| vm_spec: | ||
| GCP: | ||
| machine_type: c4-standard-8 | ||
| zone: us-central1-a | ||
| boot_disk_type: hyperdisk-balanced | ||
| boot_disk_size: 100 | ||
| sandbox_config: | ||
| type: gvisor | ||
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This tool isn't being included & therefore this comment doesn't need to be here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done, removed stale references in new commit.