No doubt, the serverless strategy saves development time. But does it also keep the expenses in check? Often portrayed as a relatively “affordable” approach, serverless, in reality, does not inherently affirm such guarantees due to multi-fold reasons.
For instance, the current price of AWS Lambda is set at $0.00001667 for every GB-second used, similar to Azure. And based on the number of functions, total memory, and the dreaded unpredictable load, it’s pretty evident how quickly the cost could get out of hand.
However, does this imply that serverless is a mediocre strategy? Certainly not! In fact, it’s part of Forbes’ curated list of top five key cloud computing trends for 2022. And with industry giants like Netflix, Coca-Cola, Amazon, Nordstrom, etc., going serverless, it’s undoubtedly a powerful strategy, but only when a business develops a profound understanding of its nuances.
In that light, this article looks at the key aspects that increase serverless costs and offers ways to optimize these expenses.
Factors That Can Cause a Price Surge in Serverless Computing
Let’s understand the three most powerful serverless offerings – Google Cloud, AWS, and Microsoft’s Azure and their respective pricing dynamics.
1. Amazon Web Services (AWS)
A serverless compute service by Amazon, AWS Lambda runs the code based on triggered events and automatically manages the required resources to complete the tasks.
Here’s what this popular serverless offering’s pricing accounts for:
- The total number of requests is set at $0.2 per 1M requests.
- The total duration of overall invocations is $0.0000097222 for every GB-second.
- Memory is configured and allocated with a price starting at $0.0000000021 for 128 MB.
2. Azure Functions
Azure Functions are also known as “trigger-based” event-driven services that offer a complete developer experience without having to worry about the infrastructure.
They are priced based on resource consumption, i.e., on a pay-as-you-go basis. Here are the other factors:
- Billed on resource consumption per second and the total number of executions. Priced at $0.000016/GB-second.
- The total number of requests, i.e., $0.20 per million executions.
3. Google Cloud Functions
With Google Cloud Functions, your developers are only in charge of the code; the rest is taken care of by Google.
Similar to Azure, Google Cloud, too, operates as per the on-demand pay-as-you-go pricing model. Additionally, it offers a long-term and a free tier option. And here’s how Google calculates the overall pricing:
- Based on the compute time, right from the functions getting requests until completion: The pricing starts from 128MB memory, 200 MHz CPU – $0.00000231 / second and upwards.
- The total number of function invocations: More than 2M invocations a month cost $0.40 per million.
- Total CPU usage and resources used in the Cloud Run.
- Outbound transfer of data priced at $0.12 per GB.
From the above metrics, it’s easy to conclude that opting for serverless could be expensive if not managed well. But it needn’t be this way. By profoundly understanding some core aspects to work around the aforementioned pricing models, you can accelerate savings while greatly benefiting from serverless.
What Can You Do to Optimize Serverless Costs?
1. Use Functions Only When Necessary
A serverless function is a business logic that mostly lasts a few seconds. And when triggered, users are charged based on the resource consumption, number of requests, and the memory used. So, a deep understanding of them and your system’s demand can help optimize costs.
Developers need to use smaller functions with less code than usual and only call them when absolutely necessary. Larger functions cause system complexity while also adding up to the otherwise avoidable costs.
2. Proactive Monitoring
Since everything outside development is at the hands of the cloud service provider, regular monitoring is vital while adopting serverless. Businesses need to track how frequently the functions or requests are run using alerts sent directly to their mailbox.
AWS, for instance, lets one set a resource quota and cap it. When this is exceeded, they are notified immediately to take appropriate action. Alerts are a fool-proof way to keep a track of the application’s health and consumption metrics.
3. Choose the Right Pricing Plan
Understanding the users’ usage patterns, the size of your workloads, budget constraints, and provisioning requirements to move to serverless can help you choose a plan that works the best.
You can initially opt for a flexible pricing plan with a pay-per-use or a pay-as-you-go model. AWS offers a savings plan for its serverless offerings that promise lower prices compared to on-demand. These plans are typically charged based on a set timeframe and usage.
Also, understand the billing model of the chosen serverless platform by talking to their support staff. Google, for instance, provides a clear set of recommendations and free cost-optimization tools for better cost control.
4. Try to Reduce Cold Starts
What are cold starts in computing? A cold start is used to define the initial call made to get a function running without the needed instance. This running time can take longer than usual as the necessary resources are yet to be allocated and therefore is of little use. So, reducing the frequency and number of cold starts can help you cut costs.
5. Outsource Your Serverless Application Development
Development in a serverless architecture does have a hidden cost in the name of hiring in-house developers. Highly skilled developers who are well versed in serverless often charge higher salaries. And to have a faultless, profitable serverless app, you need to hire an entire internal team. This is precisely where outsourcing your development to a reliable software development expert with a decent track record proves feasible.
With Zingworks, our highly experienced full-stack developers can create scalable serverless solutions while helping you drive real-time cost savings and increasing gains.
Want to unleash the power of serverless with future-ready apps? Talk to an expert today!
