ScyllaDB Simplifies Fraud Detection at Grab, SE Asia's Largest Startup

Hear Aravind Srinivasan discuss how Grab's fraud detection uses ScyllaDB's low-latency, high-performance NoSQL database. [Read the complete case study]

Transcript

One of the challenges of ride sharing or even a mobile platform, with at a scale of Grab, which is operating in pretty much all Southeast Asian countries, is near real time latency, which handles the throughput of all the customers' data. So basically, let's say if there are scammers in the world who try to scam the Grab platform, we want some platform, some technology to make sure we can flag these unfortunate rides immediately in near real time.  And that is one of the hardest problems that we have.

As part of the real time team, one of the challenges is to make sure we support our customers, the customers of my team or other engineering teams, who expect latencies in sub milliseconds. And in order to do that, we need an aggregation store which is high throughput, low latency, low overhead, low cost, maintenance as well. So that is where ScyllaDB comes in.

Low overhead is an extremely important thing. Because I always used to say, "I have a baby at home;  I don't want another baby at work."  So I don't want to keep track of DB that is sitting at work keep constant tabs. So we needed something like that. And that is where ScyllaDB comes in and it has been great so far.