In high-concurrency online transaction processing (OLTP) workloads, MySQL's active transaction list becomes a bottleneck: every transaction start, commit, and visibility check must acquire the trx sys mutex, limiting throughput and preventing full use of multiple vCPUs. PolarDB for MySQL addresses this with PolarTrans, which replaces the active transaction list with Commit Timestamp Store (CTS) — an in-memory, lock-free mechanism that tracks transaction state using commit timestamps instead of mutex-protected arrays.
This topic explains how CTS works, what it improves at each stage of the transaction lifecycle, and how performance changes after PolarTrans is enabled.
Supported versions
To enable PolarTrans through the global consistency (high-performance mode) feature, your PolarDB for MySQL Enterprise Edition cluster must meet one of the following version requirements:
Engine version 8.0.2 with a revision version of 8.0.2.2.19 or later
Engine version 8.0.1 with a revision version of 8.0.1.1.29 or later
Engine version 5.7 with a revision version of 5.7.1.0.26 or later
To check your cluster version, see the "Query the engine version" section of Engine versions.
Why active transaction lists limit performance
Mainstream relational databases such as MySQL and PostgreSQL implement Multi-Version Concurrency Control (MVCC) using an active transaction list. In high-concurrency environments, this approach creates three problems:
Cannot fully use multiple vCPUs to process concurrent transactions
Creates performance bottlenecks when many transactions compete for the
trx sys mutexDoes not adequately support read consistency, multi-point writes, and Extended Architecture (XA) transaction management in shared-nothing architectures
PolarTrans addresses these limitations by replacing the active transaction list with CTS. Instead of copying transaction state, PolarTrans determines transaction visibility based on the maximum commit timestamp of the cluster. CTS logs are maintained entirely in memory: each transaction is modulo-mapped to a specific slot in a ring buffer based on its trx_id, and each slot holds a trx pointer and a commit sequence number (CSN).
How CTS improves each transaction stage
The active transaction list requires acquiring a trx sys mutex at three critical points in the transaction lifecycle. CTS eliminates this requirement at all three stages.
| Stage | Traditional approach | PolarTrans (CTS) |
|---|---|---|
| Transaction start | Allocates a transaction ID, adds it to rw_trx_ids, updates rw_trx_set and rw_trx_list — all under trx sys mutex | Records transaction info in the CTS log, modulo-allocates a slot based on trx_id, marks the slot with an active tag — lock-free |
| Transaction commit | Queries rw_trx_ids, removes the transaction ID, updates rw_trx_set and rw_trx_list — under trx sys mutex | Assigns a commit timestamp and updates the CSN field in the CTS log only |
| Visibility determination | Copies the active transaction ID array, records the minimum and maximum transaction ID, queries the array to determine row visibility — under trx sys mutex | Compares the CSN of the read-only transaction with the trx csn field in the row record, using the system's maximum commit timestamp instead of a read view |
PolarTrans implements lock-free algorithms across most transaction logic to improve throughput in both read-write and write-only scenarios.
Performance comparison
The following test measures queries per second (QPS) before and after enabling PolarTrans.
Test environment:
| Parameter | Value |
|---|---|
| Cluster | PolarDB for MySQL 8.0, Cluster Edition |
| Specs | 88 cores, 710 GB memory |
| Test tool | Sysbench |
| Data volume | 88 tables, 12 million rows per table |
Test cases:
Results: PolarTrans significantly improves QPS in read-write and write-only scenarios.
In read-only scenarios, PolarTrans does not improve performance. Read-only transactions already benefit from read view caching, which reduces the locking overhead from copying transaction state.
Enable PolarTrans
PolarTrans is enabled by default when you enable the global consistency (high-performance mode) feature. For setup instructions, see the "Enable global consistency (high-performance mode)" section of the Overview topic.