What is a common strategy to reduce Lambda cold starts in latency-sensitive workloads?

• A. Increase Lambda memory to the maximum.
• B. Configure provisioned concurrency or keep warm; optimize code; reduce package size.
• C. Move workload to EC2.
• D. Disable VPC access.

Answer: (B)

Reducing Lambda cold starts in latency-sensitive workloads comes from having execution environments ready when requests arrive and making the startup as fast as possible. Provisioned Concurrency keeps a defined number of Lambda instances pre-warmed and initialized, so the next invocations hit ready-to-use runtimes without the usual startup delay. Keeping a small pool warm between bursts can achieve a similar effect for spikes in traffic. Alongside this, optimizing the function’s startup—such as loading fewer dependencies, refactoring to lazy-load resources, and shrinking the deployment package—reduces the work the runtime must do during initialization, further cutting latency. Increasing memory might improve overall performance but isn’t a targeted solution for cold starts and comes with higher cost. Moving to EC2 changes the execution model entirely, and disabling VPC access isn’t a practical fix for most workloads. Taken together, provisioning concurrency or keeping warm, plus code and package optimizations, best addresses cold-start latency.