Why Apache Pulsar Is Increasingly Used to Power Financial Services Applications
Today, successful financial service applications demand absolute data integrity, among other strict requirements. For example, an application like Bestpay, which processes billions of transactions every day, must implement a transaction processing pipeline that is built on a messaging platform that guarantees zero data loss. By nature, financial applications are error-intolerant for the obvious reason that an error could put our money at stake. Fortunately, there is one event streaming and messaging platform whose architecture is both accurate and robust enough to satisfy these imperatives, and that platform is Apache Pulsar.
Evolving rapidly to resolve data loss and other inadequacies in the Kafka pubsub platform, Pulsar’s unique architecture is now the ideal foundation to operate a financial app’s transaction pipeline. Beyond the demand for zero data loss, there are other absolute make-or-break necessities of a financial transaction processing app, including:
- Scalability from thousands to billions of transactions per day.
- Failover capability and automatic failure recovery.
- Fast, near-zero latency, at less than 10 ms per transaction.
- Essentially unlimited, cost-efficient storage capacity.
Here, we will survey two successful enterprise case studies in which a financial application implements Pulsar for optimal benefit. In fact, several large, progressive financial services companies now use Pulsar to handle their transaction throughput.
Let’s begin with Tencent.
Tencent Optimizes Transaction Security with Pulsar
Tencent’s internal billing system Midas processes billions of transactions at its peak. Midas handles domestic and international billing and payments, along with related functions like account management, marketing, and transaction risk management. Daily transactions range into the hundreds of millions of dollars. In view of this extraordinary liability, Tencent recognized the necessity of a transaction pipeline, which is 100% fault-proof. For reasons we will soon discover, Tencent built its golden pipeline on the Apache Pulsar event streaming platform.
While Midas’s complexity is beyond our scope here, we can imagine the size of a billing and payment system that supports billions of escrow accounts and thousands of businesses in nearly every country worldwide. Midas operates 24/7 and requires 100% effective up-time. Transactions can range from a thousand to a billion, requiring the fastest possible execution with zero error tolerance. Only Pulsar can fulfill such Promethean requirements today.
Apache Pulsar superbly satisfies the data consistency requirements of Midas’s transactions. The real-time reconciliation system serves to reduce transaction times and improve user experience during mobile payments. A message queue is also implemented to resolve transaction failures reliably. Tencent developers further optimized unique Pulsar features such as Broker, Bookie, and Zookeeper to achieve cost-effective storage via the decoupling of storage and processing that is native to Pulsar. Geo-replication, multi-tenancy, and multiple consumer modes were likewise beneficial for this global financial services enterprise application. The outcome for Bestpay is a stellar performance in Midas’s event streaming and data processing.
Orange Financial’s Bestpay Tunes Into Pulsar
Looking at a more specialized use case in which a financial services enterprise leverages the innovative architecture of Apache Pulsar, we find Orange Financial deploying Pulsar in a risk management strategy. As Vincent Xie, Chief Data Scientist and Senior Director of Orange Financial, states, “…our company leveraged Apache Pulsar to boost the efficiency of risk indicator development within Orange Financial.” Bestpay is Orange Financial’s top payment app, and as such, figures as number three globally.
Let’s explore how Bestpay reaps the benefits of Apache Pulsar.
Of China’s 700 million-plus mobile payment subscribers in 2019, more than half were registered members of Bestpay. Fifty million of these users were active last year, and the total mobile payment value was 45 trillion USD. Bestpay has a significant stake in financial services and requires a flawless transaction pipeline to maintain data integrity. Moreover, with this volume of transactions comes the enormous potential risk of fraud. A team of developers is required to hold the battlefront against diverse forms of attack, including phishing, rogue apps, trojans, identity theft, and merchant fraud. To combat these threats, Orange implements a Lambda architecture for risk detection.
The risk detection architecture consists of several layers, but the one of interest to our current scope uses Pulsar event streaming. Here, Apache Pulsar truly sparkles with multiple benefits, including:
- Simplification of custom coding.
- Improved production pipeline efficiency.
- Infrastructure and development cost reduction.
- Faster decision-making in risk management.
“Apache Pulsar greatly improved our risk management workflows…” said Xie. This led to other benefits, including easier access for non-technical, business-facing decision-makers in the risk management team. Pulsar’s BookKeeper storage nodes or bookies provide persistent message storage. Combine that with Pulsar’s layered architecture, and we have the ideal event streaming platform for our risk management system data and analysis.
Fault-Proof Financial Transactions
Pulsar’s naturally cloud-native architecture delivers event streaming ideal for financial transaction pipelines because of its focus on zero data loss and durable, decoupled storage. Thanks to the robust Pulsar platform, financial services can provide consumers and businesses with the fastest and most reliable transaction processing pipeline conceivable in today’s online payment markets. Pulsar thus becomes the backbone of the data processing component in financial apps that deploy the most resilient event streaming processes. Furthermore, Pulsar functions expose valuable methods in building machine learning and AI analytics for financial event forecasting.