4 ms·

Tiered Pricing for AWS Lambda

13 points by kiyanwang 4y ago

rmbyrro 4y ago
That will help adoption of Lambda. The primary point of serverless (financially-wise) is trading fixed costs for variable costs. But once you reach certain scale, it becomes cheaper to manage your own resources. With tiered pricing, this moment will be pushed further into the future of a project.
- arinlen 4y ago
  > That will help adoption of Lambda. The primary point of serverless (financially-wise) is trading fixed costs for variable costs. I'm not sure that's the case with AWS Lambda. It's main use is to serve as glue code between AWS services, and for that its free tier is quite roomy. I'm sure vendor lock-in plays a much bigger role than the classical cloud computing selling points.
henrydark 4y ago
Compute duration (GB-seconds) Architecture New tiered discount 0 – 6 billion x86 Same as today 6 – 15 billion x86 10% Anything over 15 billion x86 20% 0 – 7.5 billion arm64 Same as today 7.5 – 18.75 billion arm64 10% Anything over 18.75 billion arm64 20%
nateb2022 4y ago
[dupe] https://news.ycombinator.com/item?id=32349990 https://news.ycombinator.com/item?id=32349990
moralestapia 4y ago
6 billion seconds are 190 years ... This is only material for like 3 or 4 if their largest clients.
- arinlen 4y ago
  > 6 billion seconds are 190 years ... Only if each and every invocation is synchronous, sequential, with zero concurrency, and from a single lambda. Let's put it this way: a REST API made with API Gateway + AWS Lambda invokes a lambda at least once per HTTP request to a single resource. You can have multiple deployments, and multiple deployment stages.
  - moralestapia 4y ago
    I know. It was meant to put number on a more natural scale. It comes to about 16 million invocations per day. I'd guess obly a handful of their customers have this load on Lambda.
    - arinlen 4y ago
      > It comes to about 16 million invocations per day. I'd guess obly a handful of their customers have this load on Lambda. I'm not so sure about that. Without going into details, I have a project which calls a single lambda around 500 invocations/minute, give or take. That's already close to 1 million invocations per day. We're talking about 30€/month worth of lambdas.
      - moralestapia 4y ago
        That's weird, Amazon's own calculator estimates $300/mo on that. But perhaps you're somehow getting a 90% discount. Also, it's $100k for 6 billion GB/s. Unless you're spending that you won't see any savings.
        arinlen 4y ago
        > That's weird, Amazon's own calculator estimates $300/mo on that. Nah, you're just doing your math all wrong. In the very least you failed to look at how you're setting the parameters in the calculator. Give it another go and check what 500 requests per minute cost you, and play a bit with the parameters. > But perhaps you're somehow getting a 90% discount. Naw, no discount. You just didn't paid attention to your back-of-the-napkin math and exercised zero critical thinking. Lambdas are cheap like that for those who use it for what they're supposed to do: high-volume event handlers, or low-volume compute-intensive tasks. Think about it for a second: if handling 500 requests/minute really cost you $300/month, don't you think it would be far better to just provision a EC2 instance?

13 points by kiyanwang 4y ago

That will help adoption of Lambda. The primary point of serverless (financially-wise) is trading fixed costs for variable costs. But once you reach certain scale, it becomes cheaper to manage your own resources. With tiered pricing, this moment will be pushed further into the future of a project.

arinlen 4y ago
> That will help adoption of Lambda. The primary point of serverless (financially-wise) is trading fixed costs for variable costs. I'm not sure that's the case with AWS Lambda. It's main use is to serve as glue code between AWS services, and for that its free tier is quite roomy. I'm sure vendor lock-in plays a much bigger role than the classical cloud computing selling points.

henrydark 4y ago

Compute duration (GB-seconds) Architecture New tiered discount 0 – 6 billion x86 Same as today 6 – 15 billion x86 10% Anything over 15 billion x86 20% 0 – 7.5 billion arm64 Same as today 7.5 – 18.75 billion arm64 10% Anything over 18.75 billion arm64 20%

nateb2022 4y ago

[dupe] https://news.ycombinator.com/item?id=32349990 https://news.ycombinator.com/item?id=32349990

moralestapia 4y ago

6 billion seconds are 190 years ... This is only material for like 3 or 4 if their largest clients.

arinlen 4y ago
> 6 billion seconds are 190 years ... Only if each and every invocation is synchronous, sequential, with zero concurrency, and from a single lambda. Let's put it this way: a REST API made with API Gateway + AWS Lambda invokes a lambda at least once per HTTP request to a single resource. You can have multiple deployments, and multiple deployment stages.
- moralestapia 4y ago
  I know. It was meant to put number on a more natural scale. It comes to about 16 million invocations per day. I'd guess obly a handful of their customers have this load on Lambda.
  - arinlen 4y ago
    > It comes to about 16 million invocations per day. I'd guess obly a handful of their customers have this load on Lambda. I'm not so sure about that. Without going into details, I have a project which calls a single lambda around 500 invocations/minute, give or take. That's already close to 1 million invocations per day. We're talking about 30€/month worth of lambdas.
    - moralestapia 4y ago
      That's weird, Amazon's own calculator estimates $300/mo on that. But perhaps you're somehow getting a 90% discount. Also, it's $100k for 6 billion GB/s. Unless you're spending that you won't see any savings.
      - arinlen 4y ago
        > That's weird, Amazon's own calculator estimates $300/mo on that. Nah, you're just doing your math all wrong. In the very least you failed to look at how you're setting the parameters in the calculator. Give it another go and check what 500 requests per minute cost you, and play a bit with the parameters. > But perhaps you're somehow getting a 90% discount. Naw, no discount. You just didn't paid attention to your back-of-the-napkin math and exercised zero critical thinking. Lambdas are cheap like that for those who use it for what they're supposed to do: high-volume event handlers, or low-volume compute-intensive tasks. Think about it for a second: if handling 500 requests/minute really cost you $300/month, don't you think it would be far better to just provision a EC2 instance?