Given that we’re hosting Monster SCALE Summit…with tech talks on extreme-scale engineering…many of which feature our monstrously fast and scalable database, a big announcement is probably expected?
We hope this meets your super-sized expectations. Monster SCALE Summit 2025 will be featuring 60+ tech talks including:
- Just-added keynotes by Kelsey Hightower and Rachel Stephens + Adam Jacob
- Previously-teased keynotes by Avi Kivity, Martin Kleppmann + Chris Riccomini, Gwen Shapira, and Dor Laor
- Engineering talks by gamechangers like Uber, Slack, Canva, Atlassian, Wise, and Booking.com
- 14 talks by ScyllaDB users such as Medium, Disney+, ShareChat, Yieldmo, Clearview AI, and more – plus two talks by Discord
- The latest from ScyllaDB engineering: including object storage, vector search, and “ScyllaDB X Cloud”
Like other ScyllaDB-hosted conferences (e.g., P99 CONF), this conference will be free and virtual so that everyone can participate.
See the agenda and register – it’s free
Mark your calendar for March 11 and 12 because – in addition to all those great talks – you can…
- Chat directly with speakers and connect with ~20K of your peers
- Participate in some monster scale global distributed system challenges – with prizes for winners, of course
- Learn from ScyllaDB’s top experts, who are eager to answer your toughest database performance questions in our lively lounge – and preparing special interactive training courses for the occasion
- Win conference swag, sea monster plushies, book bundles, and other cool giveaways
It’s a lot. But hey, it’s Monster SCALE Summit. 🙂
Details, Details
Beyond what’s on the agenda, here’s some additional detail on a few recently-added sessions (see more in our “tiny peek” blog post)
How Discord Performs Database Upgrades at Scale
Ethan Donowitz, Senior Software Engineer, Persistence Infrastructure at Discord
Database upgrades are high-risk but high-reward. Upgrading to a newer version can make your database faster, cheaper, and more reliable; however, without thorough planning and testing, upgrades can be risky. Because databases are stateful, it is often not possible to roll back if you encounter problems after the upgrade due to backwards incompatible changes across versions. While new versions typically mean improved query latencies, changes in query planning or cache behavior across versions can cause unexpected differences in performance in places one might not expect.
Discord relies on ScyllaDB to serve millions of reads per second across many clusters, so we needed a comprehensive strategy to sufficiently de-risk upgrades to avoid impact to our users. To accomplish this, we use what we call “shadow clusters.” A shadow cluster contains roughly the same data as its corresponding cluster in production, and traffic to the primary cluster is mirrored to the shadow cluster.
Running a real production workload on a shadow cluster can expose differences in performance and resource usage across versions. When mirroring reads, we also have the ability to perform “read validations,” where the results for a query issued to the primary cluster and the shadow cluster are checked for equality. This gives us confidence that data has not been corrup
How Discord Indexes Trillions of Messages: Scaling Search Infrastructure
Vicki Niu, Senior Software Engineer at Discord
When Discord first built messages search in 2017, we designed our infrastructure to handle billions of messages sent by millions of users. As our platform grew to trillions of messages, our search system failed to keep up. We thus set out to rebuild our message search platform to meet these new scaling needs using our learnings and some new technologies.
This talk will share how we scaled Discord’s message search infrastructure using Rust, Kubernetes, and a multi-cluster Elasticsearch architecture to achieve better performance, operability, and reliability, while also enabling new search features for Discord users.
ted due to differences in behavior across versions. Testing with shadow clusters has been paramount to de-risking complicated upgrades for one of the most important pieces of infrastructure at Discord.
Route It Like It’s Hot: Scaling Payments Routing at American Express
Benjamin Cane, Distinguished Engineer at American Express
In 2023, there were over 723 billion credit card transactions. Whenever someone taps, swipes, dips, or clicks a credit or debit card, a payment switch ensures the transaction arrives safely and securely at the correct financial institution.These payment switches are the backbone of the worldwide payments ecosystem.
Join the American Express Payment Acquiring and Network team as they share their experiences from building their Global Transaction Router, which is responsible for switching and routing payments at the scale of American Express. They will explore how they’ve designed, built, and operated this Global Transaction Router to perform during record-breaking shopping holidays, ticket sales, and unexpected customer behavior.
The audience will leave with a deep understanding of the unique challenges of a payments switch (E.g., routing ISO 8583 transactions as fast as possible), some of our design choices (E.g., using containers and avoiding logging), and a deep dive into a few implementation challenges (E.g., Inefficient use of Goroutines and Channels) we found along the way.
How Yieldmo Cut Database Costs and Cloud Dependencies Fast
Todd Coleman, Chief Architect and Co-founder at Yieldmo
Yieldmo’s business relies on processing hundreds of billions of daily ad requests with subsecond latency responses. Our services initially depended on DynamoDB, and we valued its simplicity and stability. However, DynamoDB costs were becoming unsustainable, latencies were not ideal, and we sought greater flexibility in deploying services to other cloud providers.
In this session, we’ll walk you through the various options we considered to address these challenges and share why and how we ultimately moved forward with ScyllaDB’s DynamoDB-compatible API.
