ProxySQL 2.7 to 3.0: PmaControl cache-audit benchmark
The same complete PmaControl cache audit averaged 8.112 milliseconds on ProxySQL 2.7.3 and 2.015 milliseconds on 3.0.5: 75.2 percent less elapsed time, or 4.03 times faster. Client timing includes SQL, loopback communication, fetching and PHP logic. 150 timed audits per version on Ubuntu 24.04.4. Source: PmaControl pull request 6389.
PmaControl
BENCHMARK NOTES / 27 SEP 2026
ProxySQL 2.7 → 3.0
A nice improvement during PmaControl cache-audit benchmarks.
−75.2%
MEAN ELAPSED TIME PER AUDIT
4.03×
faster complete cache audits
Mean time · complete cache audit
ms/audit · lower is better
0
2
4
6
8
10
ProxySQL 2.7.3
8.112
ProxySQL 3.0.5
2.015
Supporting measurements
2.7.3
3.0.5
UNIT
Median batch mean
7.896
2.054
ms/audit
Maximum batch mean
11.914
2.919
ms/audit
Median / maximum across 15 batch means; each batch averages 10 audits.
What “ms/audit” measures
One complete cache audit = two read-only ProxySQL Admin queries.
01
SQL execution
02
Loopback + fetch
03
PHP audit logic
Client elapsed time. Connection setup, process startup, HTTP and
UI rendering are outside the timer.
The benchmark setup
Ubuntu 24.04.4 · 2 vCPU · 4 GiB RAM · PHP 8.4.26
150 timed complete audits per version
15 batches × 10 audits · 3 warmups per batch
Same audit code and SQL · 1 + 7 rows returned per audit
PHP allocated peak: 4 MiB for both versions
Scope: this cache-audit workload on a shared test VM.
These are client-observed Admin timings. Query-routing throughput
and standalone server execution time were outside this benchmark.
Source: PmaControl PR #6389 · fixed complete audit on both versions
Builds: 2.7.3-12-g50b7f85 / 3.0.5-60-g7e9e009
Nice improvement, ProxySQL.