Performance Overview

Measured benchmarks comparing aerospike-py (Rust/PyO3, native async) against the official aerospike Python client (C extension wrapped with loop.run_in_executor).

TL;DR — cumulative effect on DLRM-serving p95

324 ms → 97 ms (−70%) by stacking 3 actions: switch client → batch_read consolidation → Python 3.14t.

How fast — cumulative gain on p95

Original (official client, 3.11 GIL, gather)    ████████████████████████  324 ms   baseline
+ Replace with aerospike-py                     ██████████████             189 ms   −42%
+ gather(N) → single batch_read(mixed keys)     █████████                  126 ms   −61%
+ Python 3.14t free-threaded                    ███████                     97 ms   −70%  🔥
                                                                           ↑ 3.3× faster

Setup: FastAPI + DLRM + Aerospike CE, k6 10 VUs × 60s, 2 replicas (4 CPU / 4 GiB request).

Recommended actions (in order)

#	Action	Effect on p95
1	Replace official client → aerospike-py	−42% (Python 3.11)
2	gather(N) → single batch_read(mixed keys)	−33% more (under GIL)
3	Move runtime to Python 3.14t free-threaded	−49% more, TPS +47%
4	Keep AEROSPIKE_PY_INTERNAL_METRICS=1 on	E2E overhead ≈ 0

Pick your environment (Python 3.11 + GIL)

The thinner the surrounding stack, the larger the gap. Tail-latency advantage survives even in production-shaped workloads.

C — uvicorn + DLRM (real serving)

Full HTTP → batch_read → DLRM CPU inference → response. Closest to a real recsys serving pod.

Metric	aerospike-py	official	aerospike-py advantage
avg	118 ms	146 ms	−19%
p90	173 ms	293 ms	−41%
p95	189 ms	324 ms	−42% 🔥

→ Full numbers (Benchmarks C)

B — uvicorn ASGI only

FastAPI + uvicorn around batch_read, no model inference. The "REST API in front of a key-value lookup" shape.

Metric	aerospike-py	official	advantage
total mean	228 ms	290 ms	−21%
TPS	19.4	16.6	+17%

→ Full numbers (Benchmarks B)

A — Pure DB client (no HTTP/ML)

Python loop drives batch_read directly. Largest gap — surrounding stack is as thin as possible.

Metric	aerospike-py	official	advantage
avg mean	22 ms	108 ms	−80% 🔥
avg p99	121 ms	195 ms	−38%
TPS	374	138	+171% 🔥

→ Full numbers (Benchmarks A)

Read next

• Benchmarks — per-environment numbers, k6 raw output, server-side Prometheus metrics
• Free-Threaded Python — what changes when the GIL is removed (3.14t)
• Bottleneck Analysis — internal stage profiling, why action #2 works

To reproduce locally, see benchmark/README.md.