> For the complete documentation index, see [llms.txt](https://documentation.alluxio.io/ee-ai-en/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://documentation.alluxio.io/ee-ai-en/benchmark/s3-api/warp.md).

# Warp Benchmarks

[MinIO Warp](https://github.com/minio/warp) is a single-binary S3 performance tool for measuring GET, PUT, and mixed workload performance. Use it for **quick bucket-wide single-operation throughput numbers** when you don't need COSBench's stage machinery or httpbench's per-worker isolation.

For when to pick Warp over [COSBench](/ee-ai-en/benchmark/s3-api/cosbench.md) or [httpbench](/ee-ai-en/benchmark/s3-api/httpbench.md), see [Choosing a Benchmark Tool](/ee-ai-en/benchmark/s3-api.md#choosing-a-benchmark-tool). For reference throughput from a 6-node Warp run, see [Reference Performance Baselines](/ee-ai-en/benchmark/s3-api.md#reference-performance-baselines).

{% hint style="warning" %}
**Warp is incompatible with redirect-mode clusters.** Warp does not follow HTTP 307, and its `--list-existing` flag enumerates the entire bucket — so on a redirect-enabled 6-worker cluster, 5⁄6 of warp's GET requests hit 307 and fail. Options:

1. Disable `alluxio.worker.s3.redirect.enabled` cluster-wide before running warp (requires worker restart), or
2. Use [httpbench](/ee-ai-en/benchmark/s3-api/httpbench.md) instead — it follows 307 transparently and can target a single worker.

For which clients are affected by redirect behavior, see [Deployment Patterns](/ee-ai-en/data-access/s3-api.md#deployment-patterns).
{% endhint %}

## Prerequisites

* **Alluxio worker reachable** from the warp host on the S3 API port (default 29998).
* **Pre-loaded data** for GET tests: use `bin/alluxio job load --path /path --submit` and verify with `bin/alluxio fs check-cached /path`. Or create data via `warp put` first, as shown below.
* **Redirect disabled** (`alluxio.worker.s3.redirect.enabled=false`, the default) — see the warning above.

## Installation

```shell
# Install via Go
go install github.com/minio/warp@latest

# Or download a binary release from GitHub
```

## Usage

### Scenario: PUT Throughput (Write)

{% hint style="warning" %}
**Warp PUT compatibility**: Warp uses chunked SHA-256 payload signing by default, which Alluxio S3 does not support. This causes write failures and `InvalidTag` errors. Add `--disable-sha256-payload=true` to all `warp put` commands.
{% endhint %}

```shell
# Upload 10KB small files
warp put --host=<ALLUXIO_WORKER_IP>:29998 \
         --bucket=warp-benchmark-bucket \
         --duration=60s \
         --obj.size=10KiB \
         --concurrent=16 \
         --disable-sha256-payload=true

# Upload 1GB large files
warp put --host=<ALLUXIO_WORKER_IP>:29998 \
         --bucket=warp-benchmark-bucket \
         --region=us-east-1 \
         --obj.size=1GiB \
         --part.size=32MiB \
         --insecure \
         --concurrent=16 \
         --duration=60s \
         --disable-http-keepalive=false \
         --disable-sha256-payload=true \
         --noclear
```

### Scenario: GET Throughput (Read)

```shell
# Ensure data exists first (either via warp put above, or pre-loaded directly to S3)
warp get --host=<ALLUXIO_WORKER_IP>:29998 \
         --bucket=warp-benchmark-bucket \
         --concurrent=16 \
         --obj.size=100KiB \
         --duration=1m \
         --prefix=prefix/ \
         --list-existing
```

### Scenario: Multi-Node GET (Distributed Read)

When running warp across multiple nodes, use `--syncstart` to ensure all nodes begin at the same wall-clock time. Without it, nodes that start earlier will run uncontested for a few seconds, inflating their individual numbers and making aggregate throughput non-comparable across runs.

```shell
# Run this on every warp node simultaneously, with the same --syncstart time
warp get --host=<ALLUXIO_WORKER_IP>:29998 \
         --bucket=warp-benchmark-bucket \
         --concurrent=16 \
         --obj.size=100KiB \
         --duration=5m \
         --prefix=prefix/ \
         --list-existing \
         --syncstart=14:30  # HH:MM in 24h local time — set to ~60s from now
```

Each node writes a local `.json.zst` result file. Merge them for aggregate stats:

```shell
warp merge node1.json.zst node2.json.zst node3.json.zst
```

## Troubleshooting

* **Warp PUT fails or returns `InvalidTag`** — Warp's default chunked SHA-256 payload signing is not supported by Alluxio S3. Add `--disable-sha256-payload=true` to all `warp put` commands.
* **Warp GET returns errors or 307** — Warp does not follow HTTP 307 redirects. Ensure `alluxio.worker.s3.redirect.enabled=false` (the default), or switch to [httpbench](/ee-ai-en/benchmark/s3-api/httpbench.md).

For cross-tool troubleshooting (cache-hit verification, network bottlenecks, kernel tuning), see [Performance Tuning and Troubleshooting](/ee-ai-en/benchmark/s3-api.md#performance-tuning-and-troubleshooting) on the hub page.

## See Also

* [S3 API Benchmarks](/ee-ai-en/benchmark/s3-api.md) — overview, reference baselines, tool selection, cross-tool troubleshooting
* [COSBench Benchmarks](/ee-ai-en/benchmark/s3-api/cosbench.md) — for complex multi-stage workloads
* [httpbench Benchmarks](/ee-ai-en/benchmark/s3-api/httpbench.md) — for redirect-mode clusters or per-worker measurement
* [S3 API Setup and Configuration](/ee-ai-en/data-access/s3-api.md) — deployment patterns, endpoint setup, load balancer configuration