Recently, Alibaba Cloud announced a new milestone: PolarDB contributed record-breaking performance and cost-efficiency in TPC-C benchmark and recognized as the world's No. 1 OLTP database.
[References]
- _Alibaba Cloud News: Alibaba Cloud’s PolarDB Breaks TPC-C Benchmark World Record with Innovative ‘Three-Layer Decoupling’ Architecture >
- TPC-C Result Rankings >_
Let’s cut through the hype: When we say PolarDB achieved 2.055 billion transactions per minute (tpmC) with $0.11 per transaction (0.8 CNY/tpmC) in the TPC-C benchmark, what does this actually mean for developers?
Why TPC-C Matters (And Why Most Vendors Avoid It)
TPC-C isn’t another synthetic benchmark. It’s the database equivalent of an Ironman triathlon:
● Long-duration endurance test with ≤2% tpmC fluctuation
● 100% data correctness under hardware failure simulations
● Mixed workload: New orders (45%), payments (43%), deliveries (4%), etc.
● Atomicity checks: Rollback transactions in the event of a failure to verify consistency
The previous record holder (2023) had a performance of 814M tpmC. PolarDB’s 2.055B tpmC isn’t just faster - it’s a fundamental architectural shift.
The Engineering Breakthroughs
1. Killing Sharding Headaches
PolarDB’s Limitless cluster supports multiple RW nodes (Multi-Master) and RO nodes with near-linear performance growth. How?
● PolarTrans: Distributed transaction layer with 25μs latency via RDMA (Remote Direct Memory Access)
● ePQ: Elastic parallel query across CPU cores to enhance query performance
● SmartSSD Offload: Optimizations that help reduce Write-Ahead Logging (WAL) flush latency significantly
2. Cloud-Native, Yet Hardware-Optimized
PolarDB is designed for cloud environments while maximizing the potential of x86 architecture:
● Intel® Xeon® Processors: Utilizing advanced technologies like hardware profile-guided optimization (HW-PGO) to enhance instruction per cycle (IPC) on OLTP hotspots
● Persistent Memory (AliSCM): Capable of achieving millions of IOPS for redo log operations, significantly improving performance.
● Cost Efficiency: Database kernel optimizations to maximize computing performance per CPU core, 18K tpmC/core vs. 10K in previous record holder
Real-World Translation
This isn’t just about benchmarketing. PolarDB’s architecture solves actual pain points:
● Dynamic Scaling: From 1 to 2K nodes without downtime
● MySQL Compatibility: 100% syntax coverage
● Cost Savings: Reduce costs owing to optimized tpmC per core
Why This Matters for Developers
TPC-C is the ultimate stress test for OLTP database systems. If your database can’t pass this, your next production outage is just waiting to happen. PolarDB’s record proves that:
● Horizontal scaling can beat vertical scaling
● Cloud-native != vendor lock-in
Try It Yourself
We’re not hiding behind NDA walls:
● Tear Apart Our Code: PolarDB has an open-source version with Apache 2.0 license. Check out PolarDB on Github.
● Reproduce the Test: Check out the raw TPC-C audit report or read PolarDB performance whitepapers.
● Free Tier: Join the PolarDB Always Free Program to learn or experience PolarDB for free.
Read the blog to find out the story behind, or join the PolarDB Always Free Program and be part of PolarDB.
Top comments (0)