E-commerce platform architecture supporting multiple tenants and data processing method

Abstract

The invention discloses an e-commerce platform architecture supporting multiple tenants and a data processing method. According to the e-commerce platform architecture supporting multiple tenants, a multi-level tenant isolation mechanism, an asynchronous task arrangement engine and a promotion rule conflict self-checking algorithm are adopted; efficient allocation of hardware resources is achieved through a dynamic quota system, and resource consumption is remarkably reduced; based on a visual configuration tool, the operation process is simplified, and time and resource investment required by customer training are greatly reduced. Massive order processing in a high-concurrency scene is supported, and the throughput capacity of the system is remarkably improved; the data access efficiency is optimized by adopting a hierarchical caching strategy, and the cache utilization rate is greatly improved; and through a fine-grained field-level permission verification mechanism, the risk of cross-tenant data leakage is comprehensively reduced. The dynamic promotion rule engine supports flexible configuration, and the promotion strategy deployment time is greatly shortened; real-time synchronization and accurate management and control of inventory data are achieved, the overselling phenomenon is remarkably reduced, and the transaction stability is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce SaaS systems, and in particular to an e-commerce platform architecture supporting multiple tenants and a data processing method. Background Art

[0002] E-commerce SaaS (Software as a Service) refers to software services provided to the e-commerce industry via the internet. SaaS is an internet-based information service model in which vendors deploy application software on their own servers, allowing customers to customize and utilize these services based on their specific needs. E-commerce SaaS systems provide tenants with core functions such as product management, order processing, inventory synchronization, and promotional operations, helping companies quickly build online shopping malls and achieve collaborative management across multiple business lines. A tenant is an enterprise or individual that establishes an independent online business through the SaaS system. Each tenant has independent data space, configuration permissions, and operational strategies, such as a clothing brand's dedicated store on the platform. Inventory refers to the management of available product quantities, including both physical and virtual inventory, which must be synchronized with the sales end in real time to prevent overselling. An order is a transaction voucher generated when a user purchases a product. It contains data such as product information, payment amount, and delivery address, and is subject to full lifecycle management, including payment, fulfillment, and after-sales service. Promotions refer to marketing strategies designed to boost sales, such as discounts, limited-time discounts, and group buying. These strategies must support multiple rule combinations and conflict detection.

[0003] However, existing e-commerce SaaS systems still suffer from insufficient multi-tenant isolation, inefficient inventory and order coordination, and conflicts in dynamic promotion rules. Traditional systems often use a single database with sharded tables for isolation, resulting in a high risk of data leakage, poor cross-tenant query performance, and difficulty supporting dynamic resource quota adjustments. Existing technologies use a synchronous lock mechanism to handle inventory deductions, which can easily cause deadlocks in high-concurrency scenarios and lack a dynamic recovery strategy for pre-occupied inventory, resulting in oversell rates exceeding 5%. Traditional systems implement promotion rules through hard coding, which cannot support the superposition of multiple activities. The response time for rule conflict detection exceeds 500ms, and the cost of manual intervention is high. Therefore, it is necessary to propose an e-commerce platform architecture and data processing method that supports multi-tenancy to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide an e-commerce platform architecture and data processing method that supports multiple tenants, so as to solve the problems existing in the prior art such as insufficient multi-tenant isolation, inefficient inventory and order coordination, and conflicts in dynamic promotion rules.

[0005] In a first aspect, the present invention provides an e-commerce platform architecture supporting multiple tenants, comprising:

[0006] A highly available distributed database system, comprising a master node, read-only nodes, and standby nodes. The master node uses PostgreSQL's streaming replication technology to achieve strong data consistency and handles transaction processing and data updates. The read-only nodes use read-write separation middleware to achieve load balancing of read requests and employ weighted round-robin or minimum connection strategy to distribute requests. The standby nodes synchronize with the master node in real time and, based on a high-availability cluster management tool, achieve failover in seconds, with a recovery time objective of less than 5 seconds and zero data loss.

[0007] The multi-availability zone disaster recovery module deploys master and backup nodes across availability zones, ensures synchronization delays of less than 50 milliseconds through a dedicated network, and implements automatic failover based on heartbeat detection and distributed consensus algorithms.

[0008] Furthermore, the multi-tenant e-commerce platform architecture also includes: a PostgreSQL multi-tenant data isolation module, specifically including:

[0009] List partitioning modeling, which is used to create independent partitioned tables by tenant identifier and automatically manage partitioned tables through the PostgreSQL partition management extension;

[0010] Row-level security policy creates access control rules for each tenant, limiting access to only rows matching the current tenant identifier.

[0011] The parallel write optimization module improves the partition table write throughput by setting the maximum number of parallel working processes to 8.

[0012] Furthermore, the e-commerce platform architecture supporting multi-tenants also includes: a Redis multi-tenant cache system, specifically including:

[0013] Keyspace isolation design allocates cache keys based on the combination of tenant identifiers and business keys, and uses hash tags to force keys of the same tenant to be assigned to the same cluster shard;

[0014] The cache governance module adopts a hybrid elimination strategy and implements tenant-level API call frequency control through a current limiting module.

[0015] Furthermore, the multi-tenant e-commerce platform architecture also includes: Elasticsearch multi-tenant search architecture, specifically including:

[0016] The change data capture mechanism uses a log monitoring tool to parse change events in the database transaction log and write the event data to the message queue;

[0017] Multi-tenant index routing strategy generates independent search indexes by tenant identifier and supports cross-tenant aggregate queries and intelligent recommendations based on vector similarity.

[0018] Furthermore, the e-commerce platform architecture supporting multiple tenants also includes an inventory and order collaboration module, specifically including:

[0019] A three-level inventory control strategy uses cached atomic operations to deduct salable inventory, scheduled tasks to release overdue unpaid reserved inventory, and daily inventory reconciliation with the supplier system.

[0020] The asynchronous task orchestration engine allocates order processing flows based on task priority queues and ensures the atomicity of inventory operations through a distributed locking mechanism.

[0021] In a second aspect, the present invention provides a data processing method for an e-commerce platform supporting multiple tenants, comprising:

[0022] Data synchronization and disaster recovery processing, transaction processing and data updates based on the master node, and strong data consistency achieved through streaming replication technology;

[0023] Read requests are distributed to read-only nodes through the read-write separation middleware, and load balancing is performed using weighted round-robin or minimum connection number strategies.

[0024] The standby node synchronizes data with the primary node in real time, and implements failover in seconds based on high-availability cluster management tools, ensuring recovery time of less than 5 seconds and zero data loss.

[0025] Deploy master and backup nodes across availability zones, ensure data synchronization latency is less than 50 milliseconds through dedicated networks, and trigger automatic failover based on heartbeat detection and distributed consensus algorithms.

[0026] Furthermore, the method also includes multi-tenant data isolation processing: list partitioning of data according to tenant identifiers, automatically creating and maintaining independent partition tables through the partition management function of the database; setting row-level access control rules for each tenant, limiting them to access only data rows that match the current tenant identifier; improving the partition table write throughput through a parallel processing module, and setting the maximum number of parallel working processes to 8.

[0027] Furthermore, the method also includes multi-tenant cache management: generating cache keys in a combined format of tenant identifiers and business keys, and assigning keys of the same tenant to the same shard of the cache cluster through a hash tag mechanism; using a hybrid elimination strategy to dynamically clean up cache data, combining the least recently used and least frequently used rules, and controlling the frequency of tenant-level API calls through a current limiting function.

[0028] Furthermore, the method also includes multi-tenant search and recommendation processing: monitoring change events in the database transaction log, parsing the change data and writing it into the message queue; generating an independent search index according to the tenant identifier, supporting precise queries within the tenant, cross-tenant aggregation statistics and intelligent recommendations based on vector similarity.

[0029] Furthermore, the method also includes collaborative processing of inventory and orders: real-time deduction of salable inventory through atomic operations, regular scanning and release of overdue and unpaid reserved inventory, and daily reconciliation of inventory differences with the supplier system; allocating order processing processes based on task priority queues, and ensuring the atomicity of inventory deduction operations through a distributed locking mechanism.

[0030] The present invention has the following beneficial effects: the e-commerce platform architecture and data processing method of the present invention that supports multiple tenants realizes efficient allocation of hardware resources through a dynamic quota system, significantly reducing resource consumption; simplifies the operation process based on a visual configuration tool, greatly reducing the time and resource investment required for customer training. It supports the processing of massive orders in high-concurrency scenarios, significantly improving the system throughput; adopts a hierarchical caching strategy to optimize data access efficiency, greatly improving cache utilization; and comprehensively reduces the risk of cross-tenant data leakage through a fine-grained field-level permission verification mechanism. The dynamic promotion rule engine supports flexible configuration, greatly shortening the deployment time of promotion strategies; realizes real-time synchronization and precise control of inventory data, significantly reduces overselling, and ensures transaction stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram of the architecture of an e-commerce platform supporting multiple tenants provided by the present invention. DETAILED DESCRIPTION

[0032] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings.

[0033] See also Figure 1 , an embodiment of the present invention provides an e-commerce platform architecture supporting multiple tenants, including:

[0034] A highly available distributed database system consists of a master node, read-only nodes, and standby nodes. The master node uses PostgreSQL's streaming replication technology to achieve strong data consistency and handles transaction processing and data updates. The read-only nodes use read-write separation middleware to balance read request load and distribute requests using weighted round-robin or minimum connection strategies. The standby nodes synchronize with the master node in real time and, using high-availability cluster management tools, achieve failover in seconds, with a recovery time objective of less than 5 seconds and zero data loss. Table 1 shows the node roles and responsibilities of the highly available distributed database architecture.

[0035] Table 1 Node roles and responsibilities in a high-availability distributed database architecture

[0036]

[0037] The multi-availability zone disaster recovery module deploys master and backup nodes across availability zones, ensures synchronization delays of less than 50 milliseconds through a dedicated network, and implements automatic failover based on heartbeat detection and distributed consensus algorithms.

[0038] Specifically, the automatic failover process involves a health check, monitoring the master node's heartbeat every 2 seconds. A master node disconnection for more than 10 seconds triggers a fault determination. Based on the Raft consensus algorithm, the etcd cluster elects a new master node. Traffic is switched to the new master node, and logs from the outage are restored synchronously.

[0039] PostgreSQL multi-tenant data isolation module, specifically including:

[0040] List partitioning modeling creates independent partitioned tables based on tenant identifiers and automatically manages partitioned tables through the PostgreSQL partition management extension. Row-level security policies create access control rules for each tenant, restricting access to only rows matching the current tenant identifier. The parallel write optimization module improves partitioned table write throughput by setting the maximum number of parallel worker processes to 8.

[0041] Specifically, physical partition modeling, list partitioning (List Partitioning): Create a partition table by tenant_id Example CREATE TABLE tmall_order(tenant_id int8 NOT NULL,...) PARTITION BY LIST(tenant_id);

[0042] Automatically create a tenant partition table, and the background job listens for new tenant events.

[0043] CREATE TABLE tmall_order_tenant_001

[0044] PARTITION OF tmall_order

[0045] FOR VALUES IN(001);

[0046] Dynamic partition management: When adding new tenants, partition tables are automatically created using the pg_partman extension. Data from existing tenants is archived using the ATTACH / DETACH partition operations to avoid table locks. Performance and security enhancements are shown in Table 2.

[0047] Table 2 Performance and security enhancements

[0048]

[0049]

[0050] The Redis multi-tenant cache system specifically includes:

[0051] Keyspace isolation design allocates cache keys based on the combination of tenant identifiers and business keys, and uses hash tags to force keys of the same tenant to be assigned to the same cluster shard;

[0052] The cache management module adopts a hybrid elimination strategy and implements tenant-level API call frequency control through the current limiting module.

[0053] Specifically, the keyspace isolation design naming conventions are:

[0054] {tenant_id}:{business key} (e.g. tenant_001:product_1234_cache)

[0055] Cluster sharding strategy: Keys are distributed across 16,384 slots using the CRC16 hash algorithm. Keys belonging to the same tenant are assigned to the same shard using the hash tag ({tenant_id}). See Table 3 for the cache governance mechanism.

[0056] Table 3 Cache governance mechanism

[0057]

[0058] Elasticsearch multi-tenant search architecture, specifically including:

[0059] The change data capture mechanism uses log monitoring tools to parse change events in the database transaction log and write event data to the message queue; the multi-tenant index routing strategy generates independent search indexes based on tenant identifiers and supports cross-tenant aggregate queries and intelligent recommendations based on vector similarity.

[0060] Data synchronization mechanism, change data capture (CDC): monitor PostgreSQL WAL logs through Debezium; parse INSERT / UPDATE events and write them to Kafka; Logstash consumes messages and writes them to the corresponding ES index (index naming format is product_tenant_001)

[0061] Consistency guarantee: Adopt at-least-once delivery semantics; idempotent write ESDocument_id = source data primary key.

[0062] Table 4 Multi-tenant search solution

[0063]

[0064] Inventory and order collaboration module, specifically including:

[0065] A three-level inventory control strategy uses cached atomic operations to deduct salable inventory, scheduled tasks to release overdue unpaid reserved inventory, and daily inventory reconciliation with the supplier system.

[0066] The asynchronous task orchestration engine allocates order processing flows based on task priority queues and ensures the atomicity of inventory operations through a distributed locking mechanism.

[0067] The measured performance comparison table is shown in Table 5.

[0068] Table 5 Comparison of measured performance

[0069]

[0070] Dynamic supply chain adaptation: Solve the connection problem through the following technical breakthroughs: Protocol Description Language (DSL): Define the metadata model of the supplier interface. Sample code:

[0071]

[0072] Incremental synchronization engine: Utilizes PostgreSQL's Logical Decoding feature to capture supplier data changes and pushes them to tenants via the Kafka message queue, reducing bandwidth consumption by 78% compared to full synchronization.

[0073] Inventory and order coordination solution. Design a three-tier inventory management system: Available inventory: Real-time availability guaranteed through Redis atomic operations (DECR command deduction). Reserved inventory: A Celery scheduled task scans for timed-out, unpaid orders, releases inventory, and records it in a MySQL audit table. Physical inventory: Daily reconciliation with the supplier system, triggering alerts if discrepancies exceed 1%.

[0074] The dynamic promotion rules engine implements the following process: Operations personnel configure rules, such as "spend 200 and get 30 off." The engine compiles the rules into an abstract syntax tree (AST). Before execution, the engine calls a conflict detector, traversing the AST nodes to build a dependency graph. If a conflict is detected, such as simultaneously offering a minimum discount and a 20% discount, the optimal strategy is selected based on pre-set priorities. Key technical indicators: 5,000 rule checks per second, with a memory footprint of less than 50MB.

[0075] Page visual layout system. The core technologies include: component hot loading: encapsulating UI components as WebComponents, <dynamic-import>Implement on-demand loading; version snapshot: Generate a Git-like version hash for each modification, such as v23a8d1, and support one-click rollback; AB testing: Inject diversion logic at the Nginx layer to assign different page versions based on user ID hash.

[0076] As can be seen from the above embodiments, the present invention is based on the hash routing algorithm + dynamic schema loading technology of the tenant ID, combined with the hstore extension of PostgreSQL to realize the dynamic storage of tenant-defined fields; through the tenant resource quota manager, real-time monitoring of storage space, API call frequency and other indicators, supports elastic expansion in seconds. Based on the priority task queue of Celery, the order fulfillment process is optimized through distributed locks; a three-level inventory control strategy based on Redis atomic operations and Celery priority queues; an inventory reservation-release strategy is designed: reserved inventory is automatically recovered after 30 minutes, and eventual consistency is achieved by combining payment callbacks; a rule conflict map is constructed to detect combination conflicts such as full reduction / discounts / gifts in real time; dynamic weight configuration is supported, and the decision response time is <50ms.

[0077] Based on the above-mentioned e-commerce platform architecture supporting multiple tenants, an embodiment of the present invention further provides a data processing method for an e-commerce platform supporting multiple tenants, including:

[0078] Data synchronization and disaster recovery processing, transaction processing and data updates based on the master node, and strong data consistency achieved through streaming replication technology;

[0079] Read requests are distributed to read-only nodes through the read-write separation middleware, and load balancing is performed using weighted round-robin or minimum connection number strategies.

[0080] The standby node synchronizes data with the primary node in real time, and implements failover in seconds based on high-availability cluster management tools, ensuring recovery time of less than 5 seconds and zero data loss.

[0081] Deploy master and backup nodes across availability zones, ensure data synchronization latency is less than 50 milliseconds through dedicated networks, and trigger automatic failover based on heartbeat detection and distributed consensus algorithms.

[0082] Specifically, the method also includes multi-tenant data isolation processing: list partitioning of data according to tenant identifiers, automatically creating and maintaining independent partition tables through the database's partition management function; setting row-level access control rules for each tenant, limiting them to accessing only data rows that match the current tenant identifier; improving the partition table write throughput through a parallel processing module, and setting the maximum number of parallel working processes to 8.

[0083] Specifically, the method also includes multi-tenant cache management: generating cache keys in a combined format of tenant identifiers and business keys, and assigning keys of the same tenant to the same shard of the cache cluster through a hash tag mechanism; using a hybrid elimination strategy to dynamically clean up cache data, combining the least recently used and least frequently used rules, and controlling the frequency of tenant-level API calls through a current limiting function.

[0084] Specifically, the method also includes multi-tenant search and recommendation processing: monitoring change events in the database transaction log, parsing the change data and writing it into the message queue; generating an independent search index based on the tenant identifier, supporting precise queries within the tenant, cross-tenant aggregation statistics and intelligent recommendations based on vector similarity.

[0085] Specifically, the method also includes collaborative processing of inventory and orders: real-time deduction of salable inventory through atomic operations, regular scanning and release of overdue unpaid reserved inventory, and daily reconciliation of inventory differences with the supplier system; allocating order processing processes based on task priority queues, and ensuring the atomicity of inventory deduction operations through a distributed locking mechanism.

[0086] The above-described embodiments of the present invention do not limit the protection scope of the present invention.

Claims

1. An e-commerce platform architecture supporting multiple tenants, characterized by: include: A highly available distributed database system, including a master node, read-only nodes, and standby nodes. The master node uses PostgreSQL's streaming replication technology to achieve strong data consistency and is responsible for transaction processing and data updates. The read-only node implements read request load balancing through read-write separation middleware, and adopts weighted polling or minimum connection number strategy to distribute requests; The backup node is synchronized with the primary node in real time, and failure switching is achieved in seconds based on high-availability cluster management tools, with a recovery time objective of less than 5 seconds and zero data loss; The multi-availability zone disaster recovery module deploys master and backup nodes across availability zones, ensures synchronization delays of less than 50 milliseconds through a dedicated network, and implements automatic failover based on heartbeat detection and distributed consensus algorithms.

2. The e-commerce platform architecture supporting multiple tenants according to claim 1, characterized in that: It also includes the PostgreSQL multi-tenant data isolation module, specifically including: List partitioning modeling, which is used to create independent partitioned tables by tenant identifier and automatically manage partitioned tables through the PostgreSQL partition management extension; Row-level security policy creates access control rules for each tenant, limiting access to only rows matching the current tenant identifier. The parallel write optimization module improves the partition table write throughput by setting the maximum number of parallel working processes to 8.

3. The e-commerce platform architecture supporting multiple tenants according to claim 1, characterized in that: It also includes a Redis multi-tenant cache system, specifically including: Keyspace isolation design allocates cache keys based on the combination of tenant identifiers and business keys, and uses hash tags to force keys of the same tenant to be assigned to the same cluster shard; The cache governance module adopts a hybrid elimination strategy and implements tenant-level API call frequency control through a current limiting module.

4. The e-commerce platform architecture supporting multiple tenants according to claim 1, characterized in that: It also includes the Elasticsearch multi-tenant search architecture, including: The change data capture mechanism uses a log monitoring tool to parse change events in the database transaction log and write the event data to the message queue; Multi-tenant index routing strategy generates independent search indexes by tenant identifier and supports cross-tenant aggregate queries and intelligent recommendations based on vector similarity.

5. The e-commerce platform architecture supporting multiple tenants according to claim 1, characterized in that: It also includes inventory and order collaboration modules, including: A three-level inventory control strategy uses cached atomic operations to deduct salable inventory, scheduled tasks to release overdue unpaid reserved inventory, and daily inventory reconciliation with the supplier system. The asynchronous task orchestration engine allocates order processing flows based on task priority queues and ensures the atomicity of inventory operations through a distributed locking mechanism.

6. A data processing method for an e-commerce platform supporting multiple tenants, characterized in that: include: Data synchronization and disaster recovery processing, transaction processing and data updates based on the master node, and strong data consistency achieved through streaming replication technology; Read requests are distributed to read-only nodes through the read-write separation middleware, and load balancing is performed using weighted round-robin or minimum connection number strategies. The standby node synchronizes data with the primary node in real time, and implements failover in seconds based on high-availability cluster management tools, ensuring recovery time of less than 5 seconds and zero data loss. Deploy master and backup nodes across availability zones, ensure data synchronization latency is less than 50 milliseconds through dedicated networks, and trigger automatic failover based on heartbeat detection and distributed consensus algorithms.

7. A data processing method for an e-commerce platform supporting multiple tenants according to claim 6, characterized in that: It also includes multi-tenant data isolation processing: data is partitioned by tenant identifier, and independent partition tables are automatically created and maintained through the database's partition management function; row-level access control rules are set for each tenant, limiting their access to only data rows that match the current tenant identifier; and the parallel processing module is used to improve the partition table write throughput, setting the maximum number of parallel working processes to 8.

8. According to claim 6, a data processing method for an e-commerce platform supporting multiple tenants further includes multi-tenant cache management: generating cache keys in a combined format of tenant identifiers and business keys, and assigning keys of the same tenant to the same shard of the cache cluster through a hash tag mechanism; using a hybrid elimination strategy to dynamically clean up cache data, combining the least recently used and least frequently used rules, and controlling the frequency of tenant-level API calls through a current limiting function.

9. According to claim 6, a data processing method for an e-commerce platform supporting multiple tenants further includes multi-tenant search and recommendation processing: monitoring change events in the database transaction log, parsing the change data and writing it to the message queue; generating an independent search index based on the tenant identifier, supporting precise queries within the tenant, cross-tenant aggregation statistics, and intelligent recommendations based on vector similarity.

10. According to claim 6, a data processing method for an e-commerce platform supporting multiple tenants further includes collaborative processing of inventory and orders: real-time deduction of salable inventory through atomic operations, regular scanning and release of overdue unpaid reserved inventory, and daily reconciliation of inventory differences with the supplier system; allocating order processing procedures based on task priority queues, and ensuring the atomicity of inventory deduction operations through a distributed locking mechanism.

Images