This article is more than 1 year old

Tiers before bedtime: AWS updates Lambda pricing structure

'Time to update your spreadsheets,' expert says of 'inscrutable' changes

AWS has introduced tiered pricing for its Lambda serverless computing platform, claiming this will save on monthly costs for customers who use it to operate large workloads. It will, however, likely add to the complexity of estimating costs for many developers hoping to run a project on AWS.

Lambda is Amazon's event-driven computing service that runs user code in response to various events. Its selling point is that the platform automatically provisions and manages all the computing resources required for that code to execute, theoretically letting developers focus on their code rather than on the server hardware needed to run it.

The cloudy giant has now decided in its infinite wisdom to introduce tiered pricing for Lambda, based on a metric it uses called GB-seconds. This measures resource usage based on the number of seconds a function runs for, multiplied by the amount of RAM it has utilized.

Despite the claim about not having to worry about infrastructure, AWS is introducing different tiering structures depending on whether users opt for x86 or Arm servers. With x86, discounts kick in above 6 billion GB-seconds, with customers getting 10 percent off up to 15 billion GB-seconds, then a 20 percent discount above that.

If a developer opts for Arm servers, the discount does not start until 7.5 billion GB-seconds, with the 10 percent discount applying from there until 18.75 billion GB-seconds, and 20 percent beyond that.

AWS cites an example of a fictional company running a Lambda-based service optimized to use 2GB of memory with an average function duration of 60 seconds, which results in 75 million function invocations in one month and about 9 billion GB-seconds. If x86 servers are being used, this is 3 billion past the mark where discounting starts.

The cloud outfit claims that without tiered pricing, this would have cost this representative customer $150,015.30 in charges, while the new tiering pricing structure will see it charged $145,015.29 instead, a saving of $5,000 or about 3.3 percent.

But things are further complicated by the fact that Amazon may also charge different prices depending upon which AWS Region a customer operates it in.

Each Region is accounted for – or bucketed – separately, so a customer with Lambda workloads in multiple Regions will only get a discount for each of their workloads if each one is individually above discount threshold.

This sounds pretty complex to us, and Corey Quinn, chief cloud economist at The Duckbill Group greeted the news with a tweet proclaiming: "New inscrutable pricing complication just dropped; time to update your spreadsheets."

However, AWS says that the tiered pricing structure is applied automatically to a customer's monthly bill so no action is required on the customer's part with no changes made to the AWS Management Console, AWS CLI, or AWS SDKs.

Independent analyst Clive Longbottom agreed, saying that any added complexity only applies to efforts to predict costs, and the discount will likely be seen as worthwhile by corporate users.

"In reality, it is an automatically layered on tier that a user doesn't need to worry about," Longbottom claimed.

"For a change, users with large workloads will be able to say to their organisation that there is maximum calculated cost (as long as they can do the complex base calculations), with a level of savings that will then be able to be seen and trended against future bills."

AWS already offers a free tier of Lambda, which allows for up to a million requests per month.

Customers were also already able to take out Compute Savings Plans (CSPs), which offer large discounts in exchange for a commitment to a consistent amount of usage over a one or three-year term. Lambda will now apply tiered pricing before the CSP kicks in, according to AWS.

Better get updating those spreadsheets. ®

More about

TIP US OFF

Send us news


Other stories you might like