Leveraging Real time data in FSIs

What if the secret to sub-millisecond fraud detection isn't your microservices, but your data model? Learn why your database choice is critical for real-time performance.

#1about 5 minutes

The critical role of real-time data in modern applications

More data leads to better results in AI models and real-time decisioning, a trend seen across industries like finance, healthcare, and retail.

#2about 4 minutes

Why P99 latency matters more than the average

Average latency metrics can be misleading, while high-percentile latencies like P95 or P99 reveal the true user experience and can be approximated efficiently using bucketing.

#3about 8 minutes

Contrasting relational SQL and NoSQL data models

Relational databases use a structured schema with normalization, while NoSQL databases like Aerospike use self-describing records that favor aggregation for performance.

#4about 4 minutes

Navigating the CAP theorem in distributed systems

Distributed databases must choose between consistency (CP) and availability (AP) due to the CAP theorem, with financial services typically requiring strong consistency.

#5about 3 minutes

Defining the credit card fraud detection use case

A real-world fraud detection scenario requires loading up to 50,000 recent transactions within a strict 120-millisecond end-to-end time window.

#6about 6 minutes

Live coding a database performance testing tool

An open-source Java and Spring-based simulator is extended live to add Redis as a new database option for performance comparison.

#7about 4 minutes

Analyzing Postgres performance for fraud detection

While Postgres performs well once data is in its cache, its initial warm-up latency from disk is far too high to meet the strict real-time SLA.

#8about 6 minutes

Optimizing NoSQL data models for extreme speed

Switching from a relational-style secondary index to a bucketing model that groups transactions by day dramatically reduces latency and increases throughput in Aerospike.

#9about 2 minutes

Choosing the right database for the right job

No single database is perfect for every use case; relational databases excel at complex joins while NoSQL databases are built for high-speed, large-scale data retrieval.