Overview

SERES Group Co., Ltd.—an A-share–listed company and a China Top 500 enterprise—is a technology-driven manufacturer focused on new energy vehicles (NEVs). Since its founding in 1986, SERES has been committed to driving innovations in automotive energy and smart mobility. As the business scaled and reporting demand surged, the company’s data platform hit performance bottlenecks. Through deep collaboration with YMatrix, SERES completed a one-time migration of 2.13 TB of data in just 3 hours, achieved an average 50% improvement in multi-scenario query efficiency, and overcame long-standing challenges of slow queues, difficult migrations, and weak visualization, injecting fresh momentum into business operations.

1.Company Overview

Headquartered in Chongqing, SERES Group’s business spans the R&D, manufacturing, sales, and services of NEVs and core “three-electric” systems (battery, e-drive, e-control), with 10,000+ employees. As a leader in the NEV sector, rapid growth places higher demands on the data platform’s real-time performance and stability.

2. Challenges & Requirements

With the launch of FR direct-connect reports to Greenplum in October 2024, peak concurrent connections jumped from 200+ in September to 370 (including 150+ from FR), exposing platform bottlenecks:

Long query queues: After new reports came online, Greenplum jobs frequently queued, and query performance could not meet business needs, directly affecting operations.

Tight migration windows: Upper-layer applications must run 24/7; frequent or extended downtime for migrations was unacceptable.

Weak real-time display: During peak hours (9:00–11:00), report queries stalled with no front-end response; some reports had to be moved to MySQL/Oracle to ease peak-time pressure.

Unstable ETL jobs: Connection exceptions and I/O errors appeared during peaks, undermining data processing efficiency.

3. Solution

To address these pain points, YMatrix delivered a one-time migration from Greenplum to YMatrix:

One-shot migration with minimal business impact: Using the MatrixShift migration tool, original data in Greenplum was migrated once, avoiding repeated downtime.

One-stop data service: A single database supports both data processing and data serving, sustaining 100k+ TPS so peak-time transactional (TP) query workloads no longer require additional databases.

Stable, high-throughput ingestion: The in-house MatrixGate enables parallel writes with ingestion capacity up to 1 billion points/second.

Performance optimization: The in-house MARS3 hybrid row-column storage engine and end-to-end vectorized execution significantly improve query efficiency across diverse scenarios.

4. Outcomes

With the YMatrix-powered upgrade, SERES achieved major gains in efficiency and user experience:

Industry-leading migration efficiency:

2.13 TB migrated in 3 hours in a single pass, dramatically reducing business interruption versus multi-stage migrations.

One database, many scenarios:

YMatrix integrates processing and serving, eliminating multi-DB hops and delivering second-level report response times.

Breakthrough query performance:

• Average multi-scenario query efficiency +50%
• Manufacturing (15-minute updates): 180s → 60s (+67%)
• Marketing (30-minute updates): 240s → 120–150s (+37.5%–50%)
• EPM Sales (30-minute updates): 8–9 min → 5–6 min (+33%–44%)

Significantly stronger stability:

Peak CPU utilization dropped from >100% to ~50%; job queues and ETL errors largely disappeared, ensuring business continuity.

5. Conclusion

By leveraging the power of YMatrix’s hyper-converged database, SERES built a high-performance, highly stable data platform that underpins rapid NEV business expansion and digital transformation. Looking ahead, the two parties will deepen cooperation, explore broader data application scenarios, and jointly accelerate the automotive industry’s intelligent upgrade.