Stop 11 · Cloud Architecture PatternsTopics · 7/9

Caching Strategies

In short

The fastest request is the one that never reaches your database. A full stack app can cache in four or five places, and knowing which layer answered is half of debugging stale data.

Applying caching at multiple layers — CDN for static content, application-level caching (Redis) for computed data, and database query caching — each layer trades freshness for speed differently.

The purple blob pulls a book from a school bag next to a small bookshelf, with a library far behind.

Think of it like this

Bag, room shelf, library

Need a textbook? First you check your bag. If it's not there, the shelf in your hostel room. Then the college library. Only if the library lacks it do you order from the publisher. Each stop is slower, so you keep favourites close.

Watch it happen

Where does GET /api/products/7 get answered?

Browser

React Query

CDN

Redis

Database

1/5The browser checks its HTTP cache. With max-age=60 and a fresh copy, nothing leaves your phone.

The big ideas

What you need to know

Every cache needs an expiry

Pick a TTL per layer. Short for prices and stock, long for images and blog posts.

Plan how to clear it

When data changes, delete the Redis key, invalidate the React Query key, and purge the CDN if needed.

Never share private data

Personal responses must be Cache-Control: private or no-store, or a CDN may serve your cart to a stranger.

Cache close to the user

A hit in the browser beats a hit in Redis, which beats any database query. Cache as high up as is safe.

See it in code

Walk through the code

1app.get("/api/products/:id", async (req, res) => {2  const key = `product:${req.params.id}`;3  const cached = await redis.get(key);4  if (cached) {5    res.set("Cache-Control", "public, max-age=60, s-maxage=300");6    return res.type("json").send(cached);7  }8 9  const product = await db.product.findUnique({ where: { id: req.params.id } });10  if (!product) return res.status(404).json({ error: "Not found" });11  await redis.set(key, JSON.stringify(product), { EX: 300 });12  res.set("Cache-Control", "public, max-age=60, s-maxage=300");13  res.json(product);14});15 16// When an admin edits the product: await redis.del(`product:${id}`);

1/4Look in Redis first. On a hit, send the stored JSON without touching the database.

Cache-aside with Redis on the server, plus Cache-Control so browsers and the CDN can cache too. One route, three cache layers.

Save yourself the pain

Common mistakes

Oops

Adding a cache and never deciding how it gets cleared.

Do this instead

Write the invalidation code in the same pull request: what event clears which key, at which layer.

Oops

Putting Cache-Control: public on /api/me or /api/cart.

Do this instead

Use private or no-store for anything tied to a logged-in user.

Your mission

Find the layer that answered

~35 min

Quick check

You changed a product price, the database shows ₹499, but users still see ₹599 for a few minutes. Most likely cause?

Go deeper

1Diagram where caching could apply at 3 different layers of a hypothetical architecture.
2Learn cache invalidation strategies and why 'cache invalidation is one of the two hard problems in computer science'.
3Practice reasoning about acceptable staleness for a specific piece of cached data.