캐싱 전략: 지연 로딩 및 쓰기 관통

캐싱 전략이 중요한 이유

지연 로딩(캐시 우회 방식)의 작동 원리

# Lazy loading pattern (Python with redis-py)
def get_user(user_id, redis_client, db):
    cache_key = f'user:{user_id}'

    # 1. Check cache
    cached = redis_client.get(cache_key)
    if cached:
        return json.loads(cached)   # Cache HIT

    # 2. Cache MISS: fetch from DB
    user = db.query('SELECT * FROM users WHERE id = %s', user_id)

    # 3. Populate cache with TTL of 300 seconds
    redis_client.setex(cache_key, 300, json.dumps(user))

    return user

지연 로딩의 장점

지연 로딩의 단점

# Mitigating cache stampede with a probabilistic early expiration
# (Refresh the key before it expires to avoid simultaneous misses)
def get_with_stampede_protection(key, redis_client, db_fetch_fn, ttl=300):
    value = redis_client.get(key)
    ttl_remaining = redis_client.ttl(key)

    # Probabilistically refresh before expiry
    if value is None or (ttl_remaining < 30 and random.random() < 0.1):
        value = db_fetch_fn()
        redis_client.setex(key, ttl, json.dumps(value))
    return json.loads(value)

쓰기 관통 방식의 작동 원리

# Write-through pattern (Python)
def update_user(user_id, user_data, redis_client, db):
    # 1. Write to database FIRST
    db.execute('UPDATE users SET name=%s WHERE id=%s',
                (user_data['name'], user_id))

    # 2. Update cache immediately (write-through)
    cache_key = f'user:{user_id}'
    redis_client.setex(cache_key, 3600, json.dumps(user_data))

    return user_data

# Every read is now a cache hit for recently updated data

쓰기 관통 방식의 장점

쓰기 관통 방식의 단점

지연 로딩과 쓰기 관통 방식 결합

# Hybrid: write-through for sessions, lazy loading for product data

# Write-through for session data (always fresh, critical)
def save_session(session_id, data, redis_client, db):
    db.upsert('sessions', session_id, data)
    redis_client.setex(f'session:{session_id}', 3600, json.dumps(data))

# Lazy loading for product catalog (infrequent updates OK)
def get_product(product_id, redis_client, db):
    cached = redis_client.get(f'product:{product_id}')
    if cached:
        return json.loads(cached)
    product = db.query('SELECT * FROM products WHERE id=%s', product_id)
    redis_client.setex(f'product:{product_id}', 86400, json.dumps(product))
    return product

TTL 설계 원칙

# TTL examples by data type

# API rate limit counter: 60 seconds
redis_client.setex(f'ratelimit:{ip}', 60, count)

# User session: 30 minutes
redis_client.setex(f'session:{id}', 1800, json.dumps(session))

# Product catalog: 24 hours
redis_client.setex(f'product:{id}', 86400, json.dumps(product))

# Stock price: 10 seconds
redis_client.setex(f'price:{symbol}', 10, price)

# Static site content: 7 days
redis_client.setex(f'page:{slug}', 604800, html_content)

업데이트 시 캐시 무효화

# Delete-on-write invalidation pattern
def update_product(product_id, new_data, redis_client, db):
    # Update the database
    db.execute('UPDATE products SET ... WHERE id=%s', (product_id,))

    # Invalidate the cache key
    redis_client.delete(f'product:{product_id}')

    # Also invalidate any list/search caches that may include this product
    redis_client.delete('products:list:page:1')
    redis_client.delete(f'products:category:{new_data["category_id"]}')

    # Next read will trigger lazy loading with fresh data

쓰기 지연(쓰기 후 저장) 캐싱

# Write-behind pattern: write to cache, flush to DB asynchronously
# Application writes:
# redis_client.incr('post:42:views')      # fast, in-memory only

# Background job (runs every 60 seconds):
def flush_view_counts(redis_client, db):
    for key in redis_client.scan_iter('post:*:views'):
        count = redis_client.getdel(key)  # atomic get-and-delete
        post_id = key.split(':')[1]
        db.execute('UPDATE posts SET views = views + %s WHERE id = %s',
                   (int(count), post_id))

빠른 확인

단원 요약