Avner Sorek
Jul 30, 2017 · 3 min read

In short, I was able to move my side project — an Express.JS application — from AWS Elastic Beanstalk to Lambda+APIG. It took me less than a day and it resulted in a ~90% reduction of costs.

This could be beneficial for any non-mission critical application (or environment), and I believe that it could be a game-changer for side projects and small endeavors.

First of all, for full disclosure — the application in question is a side project and not subject to any service-level agreement or performance requirements. It is a web site called libhive. libhive scans the code in the npm registry and finds examples of packages using other packages. It is built in Node.JS with ExpressJS, and backed by DynamoDB.

libhive gets a modest 50–150 visits daily. A couple of months ago, after I decided not to pursue this project any longer, I still wanted to keep it alive — but to minimize the maintenance costs. The site was running on Elastic Beanstalk running 2 micro instances, which were on 24/7 for redundancy.

I was able to make a very quick transition to AWS Lambda and API Gateway. This was mostly due to aws-serverless-express by AWS labs, which fits ExpressJS apps to lambda almost effortlessly. Admittedly I do have some experience with Lambda and APIG, but setup did not take long, and it barely required any code changes. It took me less than a full day’s work.

The graph above shows AWS costs (in USD) across three months — April was still 100% Elastic Beanstalk, I made the change during May and June was all serverless. April’s costs were mostly Elastic Compute — specifically 18$ for two t2.micro instances and 20$ for a “classic” load balancer — all running 24/7. Those costs were reduced completely (still paying a little for some EBS volumes). During June I’ve had 631,187 requests to APIG — which came to 2.21$, and the same amount of requests to Lambda, resulting in 386,731 GB-seconds, which fall entirely under the Lambda free tier — so no cost was incurred.

Costs down with little work — too good to be true?

Of course. Moving to Lambda has its price — the service is slower, and “cold starts” do occur sometimes. This is also due to my Lambda architecture — I’m running a small monolith in a single Lambda function, that gets data from Dynamo and renders HTML views. This made the migration super-easy, but is far from an ideal serverless setup.


What this means for you?

If you’re considering this for the business you’re working for (or running) — I would not recommend moving any production applications to Lambda, especially not like this. However — if you are running multiple development/test environments on Elastic Beanstalk or EC2, you might be able to significantly reduce your costs for these non-production environments, and in larger teams/companies — you might have a lot of those, and those costs can add up.

Good news for the little guy

If you are like me — a developer with a side-project — these reduced costs could make a huge difference. The ability of having more than a static website up and running, at a tenth of the cost, can be a true game changer. Entry barriers for software products are getting lower — and I believe this is good news for everyone.


Have a similar story? A question? Would you like me to write a more technical piece about the details of the migration process? Let me know about it in the comments below. Thanks for reading.

freeCodeCamp.org

This is no longer updated. Go to https://freecodecamp.org/news instead

Avner Sorek

Written by

freeCodeCamp.org

This is no longer updated. Go to https://freecodecamp.org/news instead

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade