PayPal is the world’s largest online money transfer, billing and payments system. It owns the PayPal, Venmo, iZettle, Xoom, Braintree, and Paydiant brands. By leveraging technology to make financial services and commerce more convenient, affordable, and secure, the PayPal platform empowers over 325 million consumers and merchants in over 200 markets to join and thrive in the global economy.
About the company
The world's largest online payment system
Challenge
Scaling a fraud prevention solution with 32% annual data growth
While PayPal’s fraud rate is impressively lower than the industry average – 0.17% of revenue compared to 1.86% – that small fraction represents over $1B in annual losses. In response, PayPal built a decision platform to identify emerging fraud patterns in real time while minimizing end-user friction. The platform processed, analyzed, and merged big data with fast data, such as customer enrollment, payment, invoicing preferences, and profile data. In 2015, PayPal realized their legacy NoSQL solution was unable to keep up performance while handling massive amounts of data it was collecting and processing. PayPal began a search for a cost-effective database that could both meet their short term and long term needs to scale horizontally, ensure high performance, and uptime.
Real-time decisioning
Needed to quickly process and analyze data to identify emerging fraud patterns in under 200 milliseconds
Faster data processing
Requirement to grow from 3.5 transactions per second to 7 million transactions per second
Scaling for huge amounts of data
Data scale challenges included hundreds of petabytes
Requirements met with Aerospike
Meet SLAs 99.95%, reducing fraud exposure by 30x
Complete machine learning based fraud detection calculations in the prescribed time window otherwise they risk allowing a fraudulent transaction to occurCost effectively scale up and out
Handle extreme 32% annual growth in data and objects while keeping total cost of operations lowImprove query performance and data consistency
Strongly consistent data simplifies application logic and thus query performance simultaneously99.99% uptime
Maintain consistent high availability in a 24/7 environment necessary for digital payments
Slash data load times
While providing reliable access to fresh data — resulting in 12x reduction in re-indexing time8 million TPS
Enable more than eight million executions per second across the database environments, which includes RDBMS and NoSQL systemsResults
Scalability, growth, and cost savings — right away and over time
PayPal’s partnership with Aerospike delivered huge initial successes while providing the company a growth path to take advantage of the latest hardware and software advances. PayPal’s initial deployment with Aerospike’s patented Hybrid Memory Architecture™ (HMA) solution provided massive cost savings and huge improvements in fraud calculation, all while maintaining high performance and frictionless customer experience.
After 2015, PayPal’s fraud decision platform ballooned to over 4,000 database instances, with 100 petabytes of data, and added 32% data storage each year. In the meantime, a multiyear collaboration between Intel and Aerospike enabled optimizations for higher performance than SSDs, lower cost per GB than DRAM, higher node densities, and data indexes which persist over system restarts. PayPal took advantage of these advances and evaluated 2nd Gen Intel® Xeon® Scalable processors and Intel® Optane™ PMem to increase overall storage density and reduce DRAM costs with Aerospike. With these improvements, PayPal was able to further lower TCO and gain additional technical benefits over the initial solution, including:
30x reduction fraud exposure
Improved SLA adherence for completing fraud prevention calculations to 99.95% up from 98.5%.
8x reduction in server footprint
From 1024 servers down to 120.
3x cost reduction
$9m hardware cost savings projected from $12.5m down to $3.5m.
5x throughput improvement
From 200K to 1M transactions per second.
Testimonials
Prior to Aerospike, we were using another in-memory data store and we were running into challenges in terms of the cost of scaling…
Sai Devabhaktuni, senior director of engineering, PayPal
