Installing and maintaining database software, either on your own server or on a virtual cloud server, is a pain. You need to keep software up to date, ensure proper security, and fix any of the myriad problems that could arise at any time. If your database goes down, your application goes down with it.
Atlas, a database-as-a-service offering from MongoDB (MDB 0.85%), largely solves these problems. With Atlas, MongoDB handles many of the details, including security, backups, and performance optimization. There are no servers to manage for an Atlas customer, just a database running on one of the big cloud platforms.
Image source: MongoDB.
Going fully serverless
Until recently, Atlas came in two flavors. The low-end shared version of Atlas runs on hardware that's shared with other customers. This means that the usage patterns of other applications can affect your own application. For a production app with thousands of users, this low-cost version of Atlas isn't ideal.
Dedicated Atlas provides each customer with their own hardware to run their database, along with other advanced features not available in the shared version. This solves the main issue with the shared version, but there are still two sticking points. First, the lowest-cost dedicated plan starts a $57 per month, many times more than it would cost to simply install and run a database on a dedicated virtual server. Second, there are a lot of decisions that need to be made regarding compute power and capacity.
It may be difficult to predict how powerful a database you'll need for an application, and the requirements may change over time. The amount of compute power, memory, and storage all need to be chosen, and those decisions affect the cost of your dedicated Atlas cluster. You remove the need to manage servers with Atlas, but you still need to manage the database itself. If your app has periodic spikes in usage, you need a database powerful enough to handle those spikes. During off-times, your database will sit largely idle.
Earlier this month, MongoDB introduced a third flavor of Atlas that aims to alleviate the downsides of the dedicated Atlas plans. Serverless Atlas is still in preview, but it provides developers a way to pay for actual usage and not have to worry about deciding on specs.
The only decision a developer needs to make with serverless Atlas is which cloud provider and cloud region to use. Atlas takes it from there, dynamically scaling up and down resources as needed. Customers pay a fixed rate for each document read and document write, as well as rates for storage and data transfer.
This is similar to how other cloud-based database services work. Firestore from Alphabet's Google Cloud, for example, is fully serverless and charges for reads, writes, storage, and bandwidth in the same way. With serverless Atlas, MongoDB now competes more directly with other usage-based cloud database services.
Lowering the cost of entry
Atlas has become MongoDB's main growth engine, with sales from the database-as-a-service soaring 73% in the company's first quarter. Atlas now accounts for more than half of MongoDB's total revenue. There's no question that Atlas is a popular option for developers.
However, the high cost of a dedicated Atlas cluster may be keeping some potential customers away, especially those who only need a powerful database some of the time to deal with spikes in usage. Serverless Atlas provides an option for those customers.
MongoDB will be working with partners to help make serverless Atlas a seamless option for developers who have already embraced a serverless application architecture. The company is teaming with Vercel and Netlify, two serverless app development platforms, to integrate serverless Atlas. These partnerships will open up Atlas to new customers who may not have considered it in the past, or who assumed it wasn't a good fit.
Atlas, and services built around it, are the key to MongoDB's long-term growth story. With serverless Atlas, the company has made it easier and cheaper to use Atlas for a production application.
