42

In AWS lambda environment we can specify only the memory configuration. Is there any documentation as to what the CPU config will be for a given memory configuration?

For example,

 ApiLambda: Type: AWS::Serverless::Function Properties: Description: "This function handles the example" CodeUri: "./app/" Handler: "app.handle_request" MemorySize: 128 Timeout: 60 Runtime: python3.7

I am trying to find the minimum MemorySize for a particular number of cores. Is there any way to know the boundaries of MemorySize where the #Cores changes without bruteforcing MemorySize?

Ref:

https://aws.amazon.com/about-aws/whats-new/2020/12/aws-lambda-supports-10gb-memory-6-vcpu-cores-lambda-functions/ https://docs.aws.amazon.com/lambda/latest/dg/configuration-memory.html

Improve this question
0

2 Answers 2

Reset to default
95

The official thresholds are:

  • 1 vCPU for 1,769 MB (ref)
  • 6 vCPUs for 10,240 MB (ref)

Recent study (archive link) concludes the following for what is happening in-between:

enter image description here

Update 09/2023:

An archive link to the study quoted is used, as original link does not work.

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

The linked docu says "At 1,769 MB, a function has the equivalent of one vCPU (one vCPU-second of credits per second)." however, you show in the table 2vCPU. Can you elaborate, please?
@agoldev The table is from a third party study of that. I updated the answer with archive link as original link does not work.
10

Allocated cores by memory

In addition to @Marcin's reply, according to Chris Munns's presentation at AWS re:Invent 2020: What’s new in serverless, you can have the following allocated cores for your Lambda (which is also mentioned in AWS Documentation > Lambda > Configuring function memory (console)):

Configured Memory (MB) Allocated cores
128-1769 1
1770-3538 2
3539-5307 3
5308-7076 4
7077-8845 5
8846-10240 6

Allocated computing power by memory (CPU Cap)

Also, AWS Documentation > Lambda > Memory and computing power states that "The amount of memory also determines the amount of virtual CPU available to a function. Adding more memory proportionally increases the amount of CPU, increasing the overall computational power available. If a function is CPU-, network- or memory-bound, then changing the memory setting can dramatically improve its performance."

According to the AWS Docs > Lambda > Configuring function memory (console), "At 1,769 MB, a function has the equivalent of one vCPU (one vCPU-second of credits per second)."

Supported by my own experiments, you can utilize at maximum "N/1769 * 100"% of a vCPU where N is the configured maximum memory in MB. This is what I call CPU Cap, the maximum percentage of a CPU core (vCPU) that can be utilized at any point in a time that is bound by memory size.

Memory CPU Cap
192 MB 10.8%
256 MB 14.1%

In my experiment, I observed 38ms duration for 192MB memory (10.8% CPU Cap), and 29ms duration for 256MB memory (14.1% CPU Cap):

38ms x 10.8% CPU Cap => 410,4 unit work

410,4 unit work / 29ms => 14.1% CPU Cap

10x improvement in performance with the same cost

According to the experiment shown on the same docs, it is mentioned that "at 128 MB, the function takes 11.722 seconds on average to complete, at a cost of $0.024628 for 1,000 invocations. When the memory is increased to 1536 MB, the duration average drops to 1.465 seconds, so the cost is $0.024638. For a one-thousandth of a cent cost difference, the function has a 10-fold improvement in performance."

Memory Duration Cost
128 MB 11.722 s $0.024628
512 MB 6.678 s $0.028035
1024 MB 3.194 s $0.026830
1536 MB 1.465 s $0.024638
Improve this answer

4 Comments

Doesn't look like the ranges in the first table still hold true. The one posted by Marcin does though. Just checked it myself.
Hey @DanyloStackylo, could you please share your experiment process with us so we can cross-compare? According to our study, AWS Documentation, and AWS re:Invent presentation on YouTube support the table mentioned from my point of view.
hello, @Chooti, I first relied on your table while checking the number of cores, using Java API Runtime.getRuntime().availableProcessors(), and that didn't output as expected. Then I tried a "binary search" way to figure out at what memory level do I get 3 cores (that's what I needed for my particular experiment), and it turned out to be 3009 MB, while 3008 MB provided with only 2 cores.
Thanks for the information @DanyloStackylo, it is really weird for me to be honest. According to the AWS Documentation and Chris Munns's re:Invent 2020 Presentation the values in the table I shared should hold 🤔 I am not sure what am I missing here tbh.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.