# AWS Lambda

Sedai continuously tunes AWS Lambda memory and concurrency per function — eliminating cold starts and overprovisioning without manual configuration.

## Lambda Optimized Continuously, Not Just Once

Memory, concurrency, and cold starts require constant tuning as traffic and code change — but most teams set them once and move on. Sedai autonomously finds the optimal configuration for every function, continuously.

## Lambda's Simplicity Is Real. Its Cost Model Is Genuinely Hard to Optimize.

Memory allocation controls CPU, and CPU controls execution time, which controls cost. These relationships make Lambda difficult to tune manually — especially across hundreds of functions with different traffic profiles and runtime behaviors.

### Memory is the wrong knob when what you're tuning is CPU.

Lambda allocates CPU proportionally to memory. Teams that optimize on memory alone — without accounting for execution time tradeoffs — routinely end up paying for unused RAM or creating latency problems by under-allocating CPU.

### Static provisioned concurrency wastes money by design.

Provisioned concurrency eliminates cold starts but is billed continuously regardless of actual traffic. A fixed setting that made sense at peak will overspend during every off-peak hour until someone adjusts it.

### Reserved concurrency across hundreds of functions doesn't manage itself.

Maintaining appropriate concurrency limits across a large function fleet — as traffic patterns and deployment frequencies change — is ongoing work that SRE teams consistently deprioritize.

"Pretty much everything that we're doing is using Lambdas and Step Functions. It was a whole area that I thought I was going to have to spend a lot of resources on. But [now] the optimization of the memory and the CPU of our lambdas is something that I just have Sedai worry about. And we just see the results."

— Robert Berger, Chief Architect // Informed

## How We Help

### Memory & Performance Optimization

Sedai continuously adjusts each function's memory allocation — measuring impact on duration and cost — to find the configuration that delivers best performance at the lowest price.

### Provisioned Concurrency Management

Sedai analyzes traffic patterns to dynamically adjust provisioned concurrency throughout the day, eliminating cold starts without the cost of a static configuration.

### Reserved Concurrency Visibility

Sedai surfaces how account-level concurrency is distributed across all functions, making it practical to protect critical functions from being throttled by high-volume, non-critical ones.

## Every Function. Optimal Config. All the Time.

## Frequently Asked Questions

### How does Sedai determine the right memory setting for a function?

Sedai uses a reinforcement learning model that makes incremental adjustments to memory in the live environment, measures the resulting impact on execution duration and cost, and converges on the configuration that delivers the best performance-to-cost ratio for that specific function's behavior.

### Does Sedai's memory optimization account for the cost of execution time, not just allocation?

Yes. Because Lambda pricing is based on both memory allocated and execution duration, increasing memory can sometimes lower total cost by reducing how long a function runs. Sedai's model accounts for this relationship and optimizes for total cost, not just allocated memory.

### Will dynamic provisioned concurrency adjustments cause cold starts during transitions?

Sedai builds a predictive model from your function's traffic patterns and seasonality, adjusting concurrency ahead of anticipated demand rather than reactively. This approach is designed to maintain warm instances through traffic ramps rather than responding after cold starts occur.

### How does Sedai handle functions with unpredictable or spiky traffic?

Sedai's model accounts for traffic variability when building concurrency schedules. For highly unpredictable functions, recommendations are more conservative — maintaining sufficient baseline capacity while avoiding the cost of over-provisioning across the full day.

### Does Sedai support Lambda functions across multiple AWS accounts or regions?

Yes. Sedai supports multi-account and multi-region Lambda deployments, with concurrency visibility and optimization applied across the full account structure.