Designing a Reliable and Scalable Payment System

💰 E-commerce has grown rapidly, requiring reliable and scalable payment systems.

💳 Payment systems involve multiple parties, including payment gateways, acquiring banks, and customer banks.

🌐 There are two common ways to build a payment system: using a payment service provider (PSP) or connecting directly to banks.

💳 The payment system doesn't store card data and relies on a payment service provider (PSP) for processing transactions.

🔁 Transaction failures and inconsistencies require reconciliation in the system.

🚀 Building a reliable payment system requires considering technical concepts and handling large throughput of payment requests.

🛒 The core features of an online store payment system include generating payment events, storing payment details, and coordinating with a PSP for card processing.

🏦 The payment service updates the merchant's wallet and ledger to track account balances and financial transactions.

🔄 Communication between internal and external services can be synchronous or asynchronous.

💡 The design of a payment system needs to account for potential failures and communication issues.

💡 Synchronous communication is not suitable for most cases and asynchronous communication is preferred.

💡 Asynchronous communication, using persistent queues like Kafka, helps handle uneven traffic and prevents service overload.

💡 Implementing a reliable payment system involves addressing system failures, poison pill errors, and functional bugs.

💡 Using messaging queues like Apache Kafka can ensure transaction completion and mitigate the challenges of a payment system.

💡 Kafka is used to persist communication messages and ensure their delivery to other services.

🔄 Retry strategies are important for handling network issues and ensuring successful payment requests.

⏰ Timeouts are implemented to avoid indefinite waiting for a response.

🔍 When an order fails, it is important to determine the status of the payment and communicate it to the customer to avoid confusion.

💡 Timeouts should be set high enough to allow slower responses but also low enough to avoid waiting indefinitely.

🚫 Fallback mechanisms can be implemented to handle failures in dependent services and ensure uninterrupted payment processing.

💥 Incompatible messages, known as poison pill errors, can be saved for later debugging or discarded using a dead letter queue.

⏳ In cases where a service is down, failed transactions can be stored in a persistent queue and processed once the service recovers.

♻️ Hidden potency can be utilized to safely retry payment operations without charging the customer twice.

💳 To avoid double payments in a payment system, a unique key called a hidden potent key is used.

🔑 Hidden potent keys, often generated as UUIDs, are added to the HTTP header and serve as the ID of the payment order.

🌐 In a distributed payment system, data replication and workload distribution improve reliability and scalability.

🔒 Encryption is crucial to protect data at rest and during transmission in a payment system.

🔒 TLS and HTTPS can protect data transmitted over a network.

🔐 Access control and two-factor authentication verify user identity and restrict data access.

💾 Regularly updating software and backing up data helps prevent vulnerability exploitation and data loss.

🔑 Using long and complex passwords can enhance security against password cracking.

🔍 Monitoring data integrity allows detection of changes and malware.

✅ Reliability and fault tolerance are important for a payment system.