Industry: Mobile payment
Scale: 3.3M+ merchants, 30M transactions/day, hundreds of TB of data
Challenge: Multi-dimensional order search across 15 query dimensions at sub-second latency; permanent retention of hundreds of TB of payment data at controlled cost
Solution: LindormTable + LindormSearch + Lindorm Tunnel Service (LTS)
Outcome: Sub-second search across 15 query dimensions; millisecond real-time queries on tens of billions of records; ~USD 1M in annual storage cost savings
Customer profile
Shouqianba, operated by Shanghai Shouqianba Internet Tech Co., Ltd., is a leading mobile payment provider in China. Since its launch in December 2014—which marked the start of the one-stop payment receiving era in China's mobile payment market—Shouqianba has grown to serve more than 3.3 million merchants, processing approximately 30 million transactions per day.
Shouqianba offers merchants a full suite of services: mobile payment tools, financing, advertising, marketing management, and supply chain services. Its Saomawang smart device, which holds multiple patents in China, handles payments across all scenarios.
Challenges
Shouqianba's payment platform faced three intersecting constraints:
-
Multi-dimensional search: Merchants and operations teams need to query orders across any combination of 15 query dimensions. A rigid index structure cannot accommodate all possible query combinations.
-
Latency at scale: Queries against tens of billions of records must complete in sub-seconds. Any delay degrades the merchant experience.
-
Permanent storage at scale: The platform generates at least 75 GB of incremental data every day, on top of hundreds of terabytes of existing records. Mobile payment regulations require permanent retention, making storage cost a critical constraint.
Solution
Lindorm addresses all three constraints through a combination of its wide table engine, search engine, and real-time synchronization service.
Full-text search for multi-dimensional queries
LindormTable stores order records in a wide table structure. LindormSearch provides a full-text search layer that supports queries across any combination of query dimensions.
The two engines stay in sync through Lindorm Tunnel Service (LTS), which replicates data from LindormTable to LindormSearch in real time. This eliminates the need to modify application code when adding new query dimensions.
Hot and cold data separation for cost-efficient permanent storage
LindormTable's hot and cold data separation feature stores hot data and cold data in separate storage tiers automatically. Application code requires no changes—the engine handles data placement based on access patterns.
Hot data stays in high-performance storage for low-latency access. Cold data moves to cost-optimized storage, reducing per-GB costs while keeping the full dataset permanently accessible.
LindormTable also provides data compression optimization, further reducing the storage footprint of cold data.
Outcomes
After nearly one year in production, Shouqianba has seen measurable results across all three challenge areas:
-
Sub-second search across 15 dimensions: LindormSearch returns order search results in sub-seconds, regardless of which query dimension combination is used.
-
Millisecond real-time queries: Real-time queries on tens of billions of records return in a few milliseconds.
-
~USD 1M in annual cost savings: Combined compression and hot and cold data separation cut storage costs by approximately USD 1 million per year.
-
No O&M burden: Lindorm's fully managed service handled data migration from Shouqianba's self-managed clusters. The team no longer manages infrastructure—Lindorm provides a service level agreement (SLA) guarantee and free technical support, freeing the team to focus on its business development.
Customer feedback
We have used Lindorm clusters and services for almost one year. Lindorm experts have provided valuable suggestions and shared their extensive experience throughout the entire process from creating a solution to implementing the solution. After we raise issues, we can always receive detailed explanations and timely responses from the technical support team. The system has run smoothly since it was launched. Lindorm allows us to store large amounts of data. Lindorm can also return the results of real-time queries on tens of billions of data records in only a few milliseconds. Lindorm experts respond to our issues with patience and provide powerful support. We can save nearly USD one million per year.