Last month, Mina Naguib, Director of Operations Engineering at AdGear, and Glauber Costa, VP of Field Engineering at ScyllaDB, teamed up at the Big Data Montreal meetup to discuss Real-time Big Data at Scale, examining how Scylla helped Adwords achieve their goal of 1 million queries per second with single-digit millisecond latencies.
Click the video below to see AdGear present their real-time advertising use case along with the results of switching from Cassandra to Scylla. Glauber explains how Scylla enabled AdGear to achieve high throughput at predictably low latencies, all while keeping TCOs under tight control.
AdGear is an online ad serving platform that compiles vast amounts of consumer data that’s used by online exchanges to bid on ad placements in real-time. AdGear has very little time to enter a bid; auctions often close in less than 100 milliseconds. Cassandra simply couldn’t keep up. While it performed about 1,000,000 bids per second, the latency and tuning complexity made it costly and difficult to keep up with online exchanges.
In this talk, Mina discusses how Scylla enables AdGear to hit their extreme throughput requirements, serving the same number of requests as Cassandra with half the hardware. Originally running a 31-node Cassandra cluster, AdGear easily downsized to a 16-node Scylla cluster in which each node serves more than twice as many queries as Cassandra’s. At peak traffic, read latency fell from 21 milliseconds to less than 5 ms on Scylla.
Glauber picks up and discusses common and familiar problems with Cassandra; unpredictable latency, tuning complexity, node sprawl, and garbage collection woes. As Glauber shows, Scylla users can accomplish more with less hardware, demonstrating that a single Scylla node can achieve 1 million OPS, with 1 ms 99% latency, all while auto-tuning and scaling up and out efficiently and painlessly.
To learn more about Scylla’s close-to-the-hardware architecture and how it helped AdGear amp up their ad serving platform, check out the video above.
- Scylla Summit 2018 is around the corner. Register now!
- Learn more about Scylla from our product page.
- See what our users are saying about Scylla.
- Download Scylla. Check out our download page to run Scylla on AWS, install it locally in a Virtual Machine, or run it in Docker.