In-Memory Arrow Data Accelerator
Works with v2.0+
Create a connector instance using sample data and accelerate it using In-Memory Arrow Data Accelerator.
Requirements
Follow these steps
Step 1. Initialize a new Spice app.
spice init arrow-acceleration-qs
cd arrow-acceleration-qs
Step 2. Configure s3 dataset: copy and paste the YAML below to spicepod.yaml in the Spice app.
version: v2
kind: Spicepod
name: arrow-acceleration-qs
datasets:
- from: s3://spiceai-demo-datasets/taxi_trips/2024/
name: taxi_trips
description: taxi trips in s3
params:
file_format: parquet
Step 3. Start the Spice runtime.
spice run
Confirm in the terminal output the taxi_trips dataset has been registered:
INFO Spice.ai runtime starting...
2026-08-13T12:18:39.774775Z INFO spiced: Starting runtime v2.1.5+models.metal
2026-08-13T12:18:39.775655Z INFO runtime::init::caching: Initialized sql results cache; max size: 128.00 MiB, item ttl: 1s, hashing algorithm: XXH3, encoding: none
2026-08-13T12:18:39.775701Z INFO runtime::init::caching: Initialized search results cache; max size: 128.00 MiB, item ttl: 1s, engine: Moka
2026-08-13T12:18:39.775720Z INFO runtime::init::caching: Initialized embeddings cache; max size: 128.00 MiB, item ttl: 1s, engine: Moka
2026-08-13T12:18:39.776955Z INFO runtime::init::dataset: Dataset taxi_trips initializing...
2026-08-13T12:18:39.777562Z INFO runtime::flight: Spice Runtime Flight listening on 127.0.0.1:50051
2026-08-13T12:18:39.777944Z INFO runtime::http: Spice Runtime HTTP listening on 127.0.0.1:8090
2026-08-13T12:18:40.981333Z INFO runtime::init::dataset: Dataset taxi_trips registered (s3://spiceai-demo-datasets/taxi_trips/2024/), results cache enabled. duration_ms=0
2026-08-13T12:18:41.084720Z INFO runtime: All components are loaded. Spice runtime is ready!
Step 4. Run queries against the dataset using the Spice SQL REPL.
In a new terminal, start the Spice SQL REPL
spice sql
Query the taxi_trips dataset, observing the long query time.
select "VendorID", tpep_pickup_datetime, tpep_dropoff_datetime, passenger_count from taxi_trips limit 10;
+----------+----------------------+-----------------------+-----------------+
| VendorID | tpep_pickup_datetime | tpep_dropoff_datetime | passenger_count |
| int32 | timestamp[us] | timestamp[us] | int64 |
+----------+----------------------+-----------------------+-----------------+
| 2 | 2024-01-25T09:27:43 | 2024-01-25T09:37:09 | 1 |
| 2 | 2024-01-25T09:41:18 | 2024-01-25T09:56:19 | 1 |
| 1 | 2024-01-25T09:24:36 | 2024-01-25T10:03:59 | 1 |
| 2 | 2024-01-25T09:04:15 | 2024-01-25T09:16:53 | 1 |
| 2 | 2024-01-25T09:26:17 | 2024-01-25T09:42:50 | 1 |
| 2 | 2024-01-25T09:50:27 | 2024-01-25T10:09:28 | 1 |
| 1 | 2024-01-25T09:29:58 | 2024-01-25T09:37:14 | 1 |
| 1 | 2024-01-25T09:53:55 | 2024-01-25T10:25:37 | 1 |
| 1 | 2024-01-25T09:18:38 | 2024-01-25T09:43:35 | 1 |
| 1 | 2024-01-25T09:55:55 | 2024-01-25T10:05:55 | 1 |
+----------+----------------------+-----------------------+-----------------+
Time: 1.072190667 seconds. 10 rows.
Note: The query has no ORDER BY, so which ten rows come back varies between runs — expect different trips than shown here.
Step 5. Update the spicepod.yaml to enable In-Memory Arrow acceleration.
version: v2
kind: Spicepod
name: arrow-acceleration-qs
datasets:
- from: s3://spiceai-demo-datasets/taxi_trips/2024/
name: taxi_trips
description: taxi trips in s3
params:
file_format: parquet
acceleration:
enabled: true
Step 6. Save the changes in Spice app and observe the dataset updating and accelerating.
2026-08-13T12:19:22.038848Z INFO runtime::init::dataset: Accelerated Dataset taxi_trips updating...
2026-08-13T12:19:23.118534Z INFO runtime_table::accelerated::refresh_task: Loading data for dataset taxi_trips
2026-08-13T12:19:25.902457Z INFO runtime_table::accelerated::refresh_task: Loaded 2,964,624 rows (399.38 MiB) for dataset taxi_trips in 2s 783ms.
2026-08-13T12:19:26.977420Z INFO runtime::init::dataset: Dataset taxi_trips registered (s3://spiceai-demo-datasets/taxi_trips/2024/), acceleration (arrow), results cache enabled. duration_ms=0
Step 7. Run a query against the taxi_trips dataset again, observing the fast query time.
select "VendorID", tpep_pickup_datetime, tpep_dropoff_datetime, passenger_count from taxi_trips limit 10;
+----------+----------------------+-----------------------+-----------------+
| VendorID | tpep_pickup_datetime | tpep_dropoff_datetime | passenger_count |
| int32 | timestamp[us] | timestamp[us] | int64 |
+----------+----------------------+-----------------------+-----------------+
| 1 | 2024-01-03T10:42:24 | 2024-01-03T10:53:32 | 2 |
| 2 | 2024-01-03T10:53:30 | 2024-01-03T11:12:09 | 2 |
| 2 | 2024-01-03T10:56:10 | 2024-01-03T11:07:08 | 2 |
| 1 | 2024-01-03T10:49:00 | 2024-01-03T11:08:07 | 2 |
| 2 | 2024-01-03T10:08:43 | 2024-01-03T10:18:24 | 2 |
| 2 | 2024-01-03T10:30:50 | 2024-01-03T10:37:26 | 2 |
| 2 | 2024-01-03T10:58:54 | 2024-01-03T11:07:38 | 2 |
| 2 | 2024-01-03T10:17:19 | 2024-01-03T10:41:15 | 2 |
| 2 | 2024-01-03T10:03:09 | 2024-01-03T10:07:16 | 2 |
| 2 | 2024-01-03T10:39:29 | 2024-01-03T10:57:56 | 2 |
+----------+----------------------+-----------------------+-----------------+
Time: 0.003522625 seconds. 10 rows.
Learn more
