Optimizing Lambda for ML

123456789101112
# Deploy the AWS Lambda Power Tuning Step Functions state machine
aws serverlessrepo create-cloud-formation-change-set \
  --application-id arn:aws:serverlessrepo:us-east-1:451282441545:applications/aws-lambda-power-tuning \
  --stack-name lambda-power-tuning \
  --capabilities CAPABILITY_IAM

# Execute power tuning for your ML inference function
aws stepfunctions start-execution \
  --state-machine-arn arn:aws:states:us-east-1:YOUR_ACCOUNT:stateMachine:powerTuningStateMachine \
  --input '{
    "lambdaARN": "arn:aws:lambda:us-east-1:YOUR_ACCOUNT:function:ml-inference",
    "powerValues": [1024, 1769, 2048, 3008, 4096, 5120, 6144, 8192, 10240],

123456789101112
import os
from functools import lru_cache
from typing import Any
import logging

logger = logging.getLogger()

# Global model cache - persists across warm invocations
_model_cache: dict[str, Any] = {}

@lru_cache(maxsize=1)
def get_tokenizer():

123456789101112
{
  "lambdaARN": "arn:aws:lambda:us-east-1:123456789:function:ml-inference",
  "powerValues": [512, 1024, 1769, 2048, 3008, 4096, 5120, 6144, 8192, 10240],
  "num": 100,
  "payload": {
    "model_id": "sentiment-v2",
    "text": "Representative input text for realistic profiling",
    "include_embeddings": true
  },
  "parallelInvocation": true,
  "strategy": "cost",
  "autoOptimize": true,

123456789101112
import boto3
from botocore.config import Config
import psycopg2
from psycopg2 import pool
import time

# Global connection pool - persists across warm invocations
DB_POOL = None
S3_CLIENT = None
LAST_CONNECTION_CHECK = 0
CONNECTION_CHECK_INTERVAL = 300  # 5 minutes

123456789101112
import boto3
import json
import time
import statistics
from concurrent.futures import ThreadPoolExecutor

class LambdaMemoryOptimizer:
    def __init__(self, function_name: str):
        self.lambda_client = boto3.client('lambda')
        self.logs_client = boto3.client('logs')
        self.function_name = function_name
        self.memory_configs = [512, 1024, 1769, 2048, 3072, 4096, 5120]

123456789101112
import os
import boto3
from functools import lru_cache
from contextlib import contextmanager
import redis
import psycopg2
from psycopg2 import pool
import httpx

# Module-level initialization (runs once per execution context)
class ConnectionManager:
    _instance = None

123456789101112
import boto3
import json
import os
from datetime import datetime, timedelta
from typing import List, Dict
import asyncio
import aioboto3

class IntelligentWarmer:
    """
    Implements traffic-pattern-aware warming strategy.
    Only warms functions during expected high-traffic periods.