Middleware
Ant Group's distributed middleware products incorporate advanced industry standards and practices. They support the smooth migration and integration of traditional financial architectures, helping you reliably manage business upgrades and changes. The products also address key technical challenges in financial trading systems, such as service and data scalability, transactional consistency, disaster recovery in seconds, and elastic resource provisioning and scheduling.
Dual-mode microservices
The SOFAStack Microservice Platform manages and governs microservices using SOFAStack Microservice (SOFAStack MS) and SOFAStack Service Mesh. It supports both the SOFA framework and the Service Mesh architecture.
Microservice: Provides features for the SOFA framework, such as RPC services, dynamic configuration, throttling, and circuit breaking.
Service Mesh: Uses Service Mesh technology to support native Dubbo, Spring Cloud, and SOFA frameworks. It provides non-intrusive service management for Dubbo, Spring Cloud, and SOFA applications.
Service architecture
Benefits
High-performance distributed service framework
Provides high-performance, transparent RPC remote service calls with high scalability and fault tolerance.
Microservice administration center
Provides a set of service administration policies to ensure high-quality service operation and meet the committed service level agreement (SLA).
Highly reliable and lightweight configuration center
Provides a service for dynamically modifying configurations at application runtime and offers a graphical centralized management interface.
Active-active data centers
Supports active-active architectures in the same city and active geo-redundancy. This provides disaster recovery capabilities across different regions to ensure system availability.
Scenarios
Refactoring traditional applications into microservices |
Break down traditional financial systems into a modular, standardized, and loosely coupled microservice architecture. This approach shortens time-to-market, enables quick product listing, and helps you capture market opportunities. It improves customer service efficiency and reduces operational costs. |
Rapidly scaling high-concurrency services |
Develop Internet finance services to improve R&D efficiency and respond quickly to business changes. Scale hot spot modules to handle traffic bursts and improve user experience. This lets you offer personalized and low-cost financial products to more small and micro-sized customers. |
Active geo-redundancy across multiple data centers |
You can quickly build a scalable, high-performance, financial-grade distributed core system that features elastic scaling and active geo-redundancy. |
Distributed tracing
Distributed System Tracing (DST) is a financial-grade solution for application observability. It is designed for distributed, microservice, and cloud-native architectures. DST helps you understand complex service invocation relationships and quickly locate faults or slow nodes.
Service architecture
Benefits
Distributed cross-application tracing
Traces the complete path of each transaction. It establishes a time series of invocation times for different service interfaces, collects performance data for each service in the path, and traces transaction performance issues by service.
Multi-dimensional trace query
You can retrieve invocation chains by Trace ID based on different conditions, such as errors or timeouts. It analyzes trace information and queries invocation chain collections for various scenarios.
Multi-perspective trace display
Provides event chain diagrams, trace details, time series charts, and timelines to display system performance visually from multiple perspectives.
Correlated query of business logs
You can configure custom business logs for your application system. It automatically associates business errors and summary information to help you quickly locate problems and track business information.
Scenarios
Problem analysis and quick troubleshooting |
In distributed scenarios, service invocations are intricate, making problem analysis and troubleshooting difficult. DST can quickly locate problematic services to help resolve issues fast.
|
Optimizing application performance and architecture |
In the invocation relationship topology, you can view and analyze the number of invocations and time consumed for each application. This helps you find applications with high and low loads to make reasonable use of resources.
|
Message Queue
SOFAStack Message Queue is a distributed message middleware built on Apache RocketMQ. It provides asynchronous decoupling and peak-load shifting for distributed applications. It supports multiple message types and offers asynchronous communication that is highly reliable, highly available, and has high throughput and transactional consistency.
Service architecture
Benefits
Out-of-the-box
Supports multiple message types, such as transactional, scheduled, and ordered messages.
Supports multiple subscription patterns, such as Pub/Sub, tag filtering, clustering consumption, and broadcasting consumption.
Supports multiple protocols and languages, such as TCP/HTTP protocols and multi-language client SDKs.
Convenient O&M
Operations: Message backtracking, dead-letter queues, and monitoring and alerting.
Diagnostics: Multi-dimensional queries and end-to-end message traces.
Security: Multi-tenant resource isolation, instance-level throttling, and authorization for primary accounts and RAM users.
High performance
Low latency: Millisecond-level end-to-end latency.
Backlog capacity: Capacity for massive message accumulation.
High throughput: Processing capacity for tens of millions of concurrent requests.
Scalability: Elastic scaling based on business needs.
High availability
High availability: 99.9% service availability.
High reliability: 99.99999% data reliability.
Ensured by synchronization mechanisms, retry mechanisms, and data persistence.
Scenarios
Asynchronous decoupling |
Decouple upstream and downstream business systems using a producer-consumer model to improve business continuity. The upstream application, as the producer, puts time-consuming operations that do not require immediate (synchronous) results into the message queue. The downstream application, as the consumer, subscribes to and consumes these messages. As long as the message format remains unchanged, the sender and receiver do not need to be directly connected. A failure in one system will not affect other applications. |
Peak-load shifting |
During business peaks, such as flash sales or group buying events, user requests surge and traffic spikes. Some business systems may not be able to handle the load. After introducing a message queue, these systems act as message consumers. They can consume messages according to their own capacity, leveraging the queue's ability to handle message accumulation, and are not affected by large traffic volumes. |
Data consistency for distributed transactions |
After decoupling applications, you still need to ensure eventual data consistency. Use the transactional message feature and reliable message delivery mechanism of the message queue to achieve system decoupling while ensuring eventual data consistency. |
Task scheduling
Task Scheduler (TS) is a distributed task scheduling framework. It enables distributed processing of tasks and provides standardized, automated, visual, and centralized scheduling, monitoring, and O&M for tasks across different business systems in a financial enterprise.
Service architecture
Benefits
Supports cluster tasks
Supports sharding data by custom dimensions with no limit on the number of sharding layers. This allows for the full utilization of all machines in the cluster. For large data volumes, multi-layer scheduling tasks can process data faster. It also supports running simple single-machine tasks.
Supports graphical orchestration of task dependencies
You can orchestrate multiple tasks into a topology graph based on their execution order. The tasks run automatically without manual intervention, ensuring that many scheduled tasks are completed quickly and efficiently.
Rich task configuration items
Provides rich task configuration items, including multiple routing policies, failure retry patterns, and missed trigger handling policies. It also supports online configuration of input parameters for scheduled tasks.
Flexible trigger control
Supports triggering tasks on a schedule, periodically, by an event, or manually.
Scenarios
Batch processing in the finance industry |
This scenario requires processing specified data within a set time. The traditional method is to process data based on database sharding, where each machine handles one data shard. The disadvantage is that if the number of machines in the business cluster is greater than the number of shards, many machines are idle, leading to low resource utilization. Cluster tasks allow users to shard data by breaking down the data into index blocks. Each machine in the business cluster is assigned data to process, which provides flexibility and allows for monitoring of task execution. |
Daily cutover in the finance industry |
The daily cutover in the finance industry is a complex operation that requires multiple tasks to work together in a specific order. In a traditional implementation, the execution order is specified in the system code. If the business changes, the code must be modified for the system to work correctly, which is difficult to maintain. Using the task orchestration feature in SOFAStack Task Scheduler, you can easily adjust dependencies between tasks. This greatly improves maintainability and provides a clear view of task execution. |
API Gateway
API Gateway is an API management platform that helps enterprises centrally manage internal and external APIs. It provides high-performance, highly secure, and highly reliable communication between network-isolated systems while ensuring the security of internal systems. It is used for opening business to external partners, enabling hybrid cloud interoperability, and integrating heterogeneous applications on the corporate intranet. It helps customers innovate their business and create new scenarios.
Service architecture
Benefits
Cross-cloud/cross-domain interoperability
Supports network interoperability between hybrid clouds and multiple business domains, enabling unified API management.
Weak network optimization
Optimizes data transmission communication protocols for weak networks based on the characteristics of mobile devices to improve network communication quality and efficiency.
Heterogeneous integration
Supports seamless integration between business systems that are implemented with different languages and technology stacks.
Support for multiple client access types
Supports mobile clients (iOS and Android) and non-mobile miniapps (H5, Web, and backend applications).
End-to-end full-path tracing
Supports end-to-end tracing for requests from the client to all nodes of the backend service.
Superior security capabilities
Supports multiple identity authentication and data encryption methods, such as CORS and SSL, to ensure the security of data transmission.
Scenarios
Frontend-backend separation scenario |
|
Heterogeneous system integration scenario |
|
Hybrid cloud scenario |
|
Data access proxy
Open Database Proxy (ODP) can address data access bottlenecks and database disaster recovery issues under massive request loads. It provides an online distributed database service with horizontal splitting, smooth scaling, and read/write splitting. It offers a lightweight, low-resource, high-performance, and highly available solution for massive data access.
Service architecture
Benefits
Sharding
Supports sharding for RDS, OceanBase, and MySQL. After you create a distributed database and select a shard key, ODP can generate sharding rules based on the shard key to implement horizontal data splitting.
Transparent read/write splitting
You can implement read/write splitting using ODP's read-only instances or MySQL standby machines. This feature handles details such as transactions, read-only instance or standby machine failures, and specified primary/standby access. It is non-intrusive to the application. You can complete read/write splitting operations in the ODP console.
Smooth scale-out for data storage
When data storage capacity and access volume bottlenecks occur, ODP supports online storage capacity expansion. The scale-out process does not require application modification, and you can track the progress visually.
Service upgrade or downgrade
You can elastically scale the service capabilities of ODP instances by changing the resource count.
Globally unique numeric sequence
ODP supports a distributed, globally unique, and orderly incrementing numeric sequence. This meets business requirements for primary or unique keys in specific scenarios when you use a distributed database.
Database account permission system
ODP supports an account and permission system similar to that of a standalone MySQL instance, which ensures the operational security of accounts used by different roles.
Distributed transaction support
ODP, when combined with the middleware distributed transaction suite, can support distributed transactions to ensure data consistency in distributed databases.
Scenarios
Massive data reads and writes |
Provides a flexible data sharding mechanism with low code intrusion. You can easily implement horizontal data splitting and scale-out to solve single-machine database bottlenecks. |
Financial-grade data disaster recovery |
Based on Ant Group's years of internal experience with financial-grade data disaster recovery scenarios, it provides a variety of data center-level data disaster recovery solutions for different business scenarios. This ensures data stability and business continuity. |
Database traffic distribution |
Provides a rule-based traffic distribution mechanism. It meets the needs of high-query businesses by deploying multiple database instances, such as one primary for writes and multiple secondaries for reads. |
Distributed transaction
Distributed Transaction-eXtended (DTX) is a financial-grade distributed transaction middleware. It is used to ensure the eventual consistency of business activities in large-scale distributed environments. Within Ant Group, it is widely used in core financial transaction paths, such as trading, transfers, and red envelopes, to support fund operations for hundreds of millions of users. DTX can be used with service frameworks (such as SOFABoot, Spring Cloud, and Dubbo), data sources (such as ODP, RDS, MySQL, and OceanBase), and other middleware products like message queues to easily implement various combinations, such as service-path-level transactions, cross-database transactions, and message transactions.
Service architecture
Benefits
Supports multiple access patterns
SOFAStack distributed transaction has evolved with business diversity to include multiple access patterns. It flexibly adapts to various business scenarios in financial technology.
TCC pattern: TCC (Try-Confirm-Cancel) is a high-performance distributed transaction access solution. This pattern provides more flexibility and can meet almost any transaction scenario. The TCC pattern supports scenarios such as custom compensation-based transactions, custom resource reservation-based transactions, and message transactions. You can intervene in the two-phase commit protocol to implement custom optimizations and special functions for specific scenarios.
FMT pattern: To address the usability issues of the TCC pattern, the distributed transaction service introduced the Framework-Managed Transactions (FMT) pattern. FMT is a non-intrusive distributed transaction solution that solves the usability problem of distributed transactions. Its main features are ease of use, fast access, and no intrusion into business code.
Saga pattern: Saga is a compensation protocol. In the Saga pattern, a distributed transaction has multiple participants. Each participant is a reversal and compensation service. You need to implement its forward operation and reverse rollback operation according to the business scenario.
Financial-grade quality assurance
Financial-grade disaster recovery assurance: Provides multiple patterns and levels of disaster recovery capabilities, including same-city and remote-region options. It ensures customer business continuity using the highest industry standards.
Unparalleled performance: Compared to the traditional two-phase commit protocol, it reduces lock holding time and significantly improves performance. The unique Performance Bursting Mode can greatly increase throughput. It supported 544,000 transactions per second during the 2019 Double 11 shopping festival.
Simple and easy to access: Ant Group's years of practical experience have resulted in a product with fast and flexible access capabilities, making it easy to use and maintain.
Compatibility assurance: A distributed transaction is an abstract concept based on the service layer and is independent of the underlying transaction implementation. This means that within the scope of a distributed transaction, various data stores can be integrated. These include relational databases such as MySQL or Oracle, key-value stores such as MemCache, and column-store databases such as HBase. This is possible as long as their operations are wrapped as participants in a distributed transaction.
Scenarios
Payments and transfers |
Common business scenarios in the finance industry, such as payments, transfers, and accounting, have high throughput requirements. The excellent performance of SOFAStack distributed transaction in various major sales promotions proves that performance will not be a bottleneck. |
Wealth management |
These scenarios often involve large amounts of money, so the stability requirements for the product are very high. SOFAStack distributed transaction has financial-grade quality and can ensure business continuity and stability. |
Insurance and regulatory reporting |
Multiple participants and high business complexity are typical features of this type of business. After more than ten years of evolution, SOFAStack distributed transaction is flexible enough to handle various scenarios, meet transactional consistency requirements, and ensure perfect integration with all kinds of businesses. |