‼️ Storage Bloat with Poor Compression: Yulu’s IoT data exceeded 18TB monthly. MySQL lacked efficient compression and smart storage handling, causing high IOPS demand, bloated AWS bills, and difficulty retaining historical data.
‼️ Write Bottlenecks During Traffic Spikes: RDS MySQL struggled with the volume of incoming IoT writes. During peak loads, IOPS throttling led to high write latency and occasional query failures.
‼️ Downtime from Schema Changes (DDL): DDL operations like adding columns in large MySQL tables led to locking and application slowdowns. Managing schema changes at scale was painful and risky.
‼️Low Availability with Single-Master Setup: A single-master database meant any failure caused downtime. Lack of automatic failover hurt service availability and user trust.
‼️ Scaling Limitations: The vertically scaled architecture couldn’t independently scale reads and writes. Replication lag and rigid scaling slowed performance during high demand.
→ Cut Storage & IOPS Costs: Reduce storage footprint and AWS expenses.
→ Improve Write Scalability: Handle high-volume IoT data without throttling.
→ Reduce DDL Downtime: Avoid app impact during schema changes.
→ Ensure High Availability: Eliminate single points of failure.
→ Adopt Scalable Architecture: Enable future growth with distributed design.
Our team began by auditing Yulu’s schema, indexes, and query patterns. We also studied traffic loads and application behavior to identify bottlenecks.
Based on Yulu’s scale and growth plans, we designed a production-grade TiDB architecture:
This ensured separation of compute and storage, fault tolerance, and distributed write/read operations.
To avoid long downtime, we used TiDB’s Data Migration (DM) tool. It allowed real-time replication from MySQL, letting Yulu continue operations on MySQL while data synced in the background.
TiDB’s TiKV storage engine uses Log-Structured Merge Trees, which optimize both disk I/O and compression. This led to a huge reduction in final storage size post-migration — from 18TB to 3.4TB.
We leveraged TiDB’s online DDL capabilities so schema changes could happen without locking tables — a major operational win over MySQL.
TiDB’s automatic failover ensured availability, even if a node failed. This removed dependency on a single-master and minimized impact during maintenance or failure.
We optimized queries for distributed execution, reviewed resource usage, and established alerts and monitoring for proactive operations.
Yulu’s disk usage dropped from 18TB to 3.4TB, saving costs on both storage and IOPS. The use of LSM trees in TiKV was critical in this outcome.
DDL operations were now seamless and did not impact application availability. Developers could push schema updates with confidence.
TiDB’s distributed architecture allowed writes to scale horizontally. The system handled traffic surges gracefully without hitting IOPS or compute bottlenecks.
Automatic failovers and a multi-node setup removed single points of failure. This significantly increased reliability and reduced support escalations.
Thanks to live data replication during migration, Yulu experienced negligible downtime. The cutover to TiDB was smooth and controlled.
Yulu is now positioned to explore Hybrid Transactional and Analytical Processing (HTAP), combining real-time operational workloads with in-depth analytical queries — all on TiDB.
At the heart of every mobility revolution is data — and for Yulu, ensuring that data remains fast, scalable, and reliable was a non-negotiable need. With over 72% reduction in storage and dramatic improvements in write scalability, Yulu’s transformation wasn’t just technical — it was strategic. By partnering with Mydbops, they didn’t just migrate databases — they unlocked new possibilities, better cost efficiency, and peace of mind for their tech team.
This isn’t just a story about TiDB. It’s a story about choosing a partner who understands the urgency of performance, the weight of reliability, and the power of proactive architecture. If you're facing performance bottlenecks, spiraling cloud bills, or scalability roadblocks — you don’t have to solve them alone. We’re here to help.
Talk to a Database Expert Today!
.svg)
