AWS Lambda Support
AWS Lambda power tuning: the right memory for every function.
Lambda bills memory multiplied by duration, and memory also decides how much CPU a function gets. The cheapest setting is rarely the lowest memory, and the fastest is rarely the best value. We measure each function and pick the configuration that fits your goal.
When You Need This
The situations where this work pays for itself.
Signs it is time
- Functions are still at the default 128 MB or at a round number nobody chose
- Your Lambda bill is growing and you do not know which functions drive it
- p95 latency is high on CPU-bound work such as parsing, image handling or encryption
- You plan to move to Graviton (arm64) and want proof it helps
- A runtime, dependency or code change may have shifted the cost curve
What you get
- Before and after cost per million invocations for each tuned function
- Recommended memory size and architecture per function, with the evidence behind it
- Infrastructure as code changes ready to merge
- A reusable tuning pipeline and a short runbook for your team
How we do it
A clear, safe process you can follow along with.
1. Find the functions worth tuning
We rank functions by invocations multiplied by duration and memory using Cost Explorer and CloudWatch, so effort goes where the money is.
2. Make functions safe to test
Power tuning really invokes your function. We prepare representative test payloads and use an alias or staging copy so tests never write production data, send email or charge customers.
3. Deploy the tuning state machine
We deploy the open-source AWS Lambda Power Tuning Step Functions workflow into your account with infrastructure as code, with least-privilege access to the functions under test.
4. Run it across memory sizes and architectures
Each function runs many times at several memory values, in parallel, on x86_64 and arm64, with enough invocations to smooth out noise. We test warm and cold starts separately.
5. Read the curve, not just the winner
We find the point where extra memory stops cutting duration, then pick for lowest cost, fastest response or a balance, depending on the function's role.
6. Roll out and verify
We apply the change with a weighted alias rollout, then compare real p95 latency, error rate and cost per million invocations before and after.
7. Keep it tuned
We add a repeatable tuning job to CI so changes to code, dependencies or runtime are re-tested instead of forgotten.
Mistakes we help you avoid
Common ways this goes wrong, and what we do instead.
Tuning with unrepresentative payloads
A tiny test event hides the real duration. We use payloads that match production size and shape.
Ignoring cold starts
Averages over warm invocations flatter a function. We measure initialization separately and tune for your latency target.
Forgetting downstream limits
A faster function sends more requests per second to your database or API. We check connection and rate limits before raising memory.
Tuning functions with side effects
Tests that write, send or charge are dangerous. We use idempotent test modes or isolated copies.
Tuning once and never again
Code and dependency changes move the optimum. We schedule re-tests.
Frequently Asked Questions
What is AWS Lambda Power Tuning?
It is an open-source tool that runs a Lambda function at different memory settings and reports how duration and cost change, so you can choose the best configuration with data.
Does more memory always cost more?
No. More memory also adds CPU, so a function can finish much faster and cost the same or less overall, up to a point. Past that point cost only rises.
Is it safe to run on production functions?
It invokes the function for real, so we only run it with safe test payloads or on an isolated copy, and never on functions with side effects we cannot contain.
Should we move Lambda to Graviton?
Often yes, because arm64 has a lower price per GB-second, but native dependencies must be compatible. We test both architectures with the same payloads.
Want Help With AWS Lambda power tuning: the right memory for every function?
Schedule an appointment. Tell us about your environment and we will outline the safest, highest-value approach.
Schedule Appointment › No commitment required