Enhancing Scalability and Reducing Costs by 40% Using Amazon Keyspaces with Adobe

LES facilitates seamless collaboration across Adobe’s suite of creative applications, helping multiple users work on the same document simultaneously. To power these capabilities, the service requires a robust, highly available database capable of handling ephemeral data—small, temporary changes that are distributed among collaborators.

Based in Hamburg, Germany, the Adobe Content Platform and Collaboration (ACPC) team oversees LES. Initially, the team relied on a self-hosted Apache Cassandra database cluster that ran on Amazon Elastic Compute Cloud (Amazon EC2) instances, which provide secure and resizable compute capacity. But to save time and increase efficiency, the team wanted to reduce operational overhead (for example, the time it takes to scale the database), minimize manual work, improve disaster recovery, and simplify data deletion.

When the ACPC team learned about Amazon Keyspaces, it saw an opportunity to simplify database management while gaining additional benefits. “After some calculations, we also realized that we could gain cost savings from this migration,” says Matthias Schulz, senior computer scientist and team lead at Adobe.

The migration was complex because Adobe needed to extract large volumes of data from the Cassandra cluster while maintaining data integrity and minimizing service disruptions. To avoid downtime for LES users, the ACPC team carried out the actual migration using a dual-write strategy.

The team wrote new data to both the existing Cassandra cluster and the new Amazon Keyspaces tables simultaneously. It also used Apache Spark–based jobs on AWS Glue, a serverless data integration service, to efficiently export historical data in bulk from Cassandra and import it into Amazon Keyspaces.

Throughout the process, the ACPC team met with members from the AWS team weekly to address challenges and optimize the migration approach. In 6 months, Adobe completed the migration while maintaining virtually zero downtime for its customers.

LES now uses Amazon Keyspaces as the primary database, which has brought several benefits to the ACPC team. Amazon Keyspaces automatically scales as needed to accommodate growing data volumes and user loads. The ability to scale and provision resources quickly has empowered the LES team to speed up the releases of new features.

“If we wanted to ship a new feature, we needed to provision database tables across our environments. This would take days using Cassandra,” says Schulz. “Now, using Amazon Keyspaces, we can implement database tables for new features within hours.”

Using Amazon Keyspaces, Adobe has minimized many of the operational tasks that are needed to keep Cassandra clusters running smoothly. For example, Amazon Keyspaces automatically deletes data without leaving deprecated records (also known as tombstones) behind, removing the need for manual intervention. This automation has reduced operational complexity and freed up engineers for more innovative work and value-added tasks instead of database management.

Additionally, disk and compute resources are decoupled in Amazon Keyspaces, which provides greater flexibility and efficiency in resource usage. The ACPC team can now allocate resources according to actual usage patterns, which has led to reduced manual effort in capacity planning and cost savings of up to 40 percent.

Using Amazon Keyspaces, Adobe is also taking advantage of point-in-time recovery. With this feature, the ACPC team can quickly restore database states without manual work. The team can perform disaster-recovery tests more frequently and with greater confidence, improving system resilience.

“The reduced disaster-recovery effort is a huge win,” says Schulz. “Although this process was automatic with Cassandra, it still took time because we needed to restore the backup into a new cluster manually. Now, we do not have to worry about complex procedures when we replicate and restore data.”

With the support of the AWS team and careful planning, the ACPC team completed a complex migration without affecting the customer experience. The success of this migration has sparked interest among other Adobe teams that use Cassandra. Several teams are now considering similar migrations to Amazon Keyspaces, using the ACPC team’s experience as a blueprint.

Looking ahead, the ACPC team is exploring ways to create new, independently scalable tables and improve data structures on the cloud. This agility will help enhance the performance of Adobe’s collaborative services even further.

“People are often afraid of migrations because they involve tedious work and a lot of risk,” says Schulz. “We had a very good experience working alongside the AWS team, and it’s good for teams across Adobe to see that others have succeeded with a project like this.”