3.4. Optimizing Application Performance & Cost

💡 First Principle: Performance and cost are two sides of the same coin in serverless — optimizing one almost always improves the other.

Why does Lambda cold start matter? Because every 100ms of initialization latency costs you both user experience and money (you pay for that time). Without understanding memory-to-CPU ratios, connection pooling, and caching strategies, developers deploy applications that are simultaneously slow and expensive. Consider a Lambda function allocated 128 MB of memory processing image uploads — it takes 8 seconds per image. Bumping to 1024 MB cuts that to 1 second and actually costs less because you pay per ms.

Unlike traditional infrastructure where you over-provision for peak load, serverless lets you pay per-request — but only if you've optimized per-request cost. A poorly written Lambda that runs for 10 seconds costs 100x more than an optimized version running for 100ms.