update

Author: andygrove

tpch/README.md

@@ -19,24 +19,26 @@
 
 # Benchmarking DataFusion Ray on Kubernetes
 
-This is a rough guide to deploying and benchmarking DataFusion Ray on Kubernetes.
+This is a rough guide to deploying and benchmarking DataFusion Ray on Kubernetes as part of the development process.
 
-set up new venv
+## Building Wheels
 
-```shell
-python3 -m venv venv
-source venv/bin/activate
-pip3 install maturin
-pip3 install ray
-pip3 install ray[default]
-```
+Follow the instructions in the [contributor guide] to set up a development environment and then build the project 
+using the following command.
 
-Build the project.
+[contributor guide]: ../docs/contributing.md
 
 ```shell
 maturin build --strip
 ```
 
+## Create a Ray Cluster
+
+Create a `datafusion-ray.yaml` file using the following template. It is important that the Ray Docker image uses the 
+same Python version that was used to build the wheels. This example yaml assumes that the TPC-H data files are 
+available locally on each node in the cluster at the path `/mnt/bigdata`. If the data is stored on object storage then 
+the `volume` and `volumeMount` sections can be removed.
+
 ```yaml
 apiVersion: ray.io/v1alpha1
 kind: RayCluster
@@ -93,11 +95,14 @@ spec:
                 claimName: ray-pvc
 ```
 
+Run the following command to create the cluster:
+
 ```shell
 kubectl apply -f datafusion-ray.yaml
 ```
 
-set up port forwarding on head node 8265
+Once the cluster is running, set up port forwarding on port 8265 on the head node and then run the following 
+command to run the benchmarks.
 
 ```shell
 ray job submit --address='http://localhost:8265' \

tpch/datafusion-ray-tpch-1739638664197.json

@@ -0,0 +1,14 @@
+{
+    "engine": "datafusion-ray",
+    "benchmark": "tpch",
+    "settings": {
+        "concurrency": 8,
+        "batch_size": 8182,
+        "prefetch_buffer_size": 0,
+        "partitions_per_worker": null
+    },
+    "data_path": "file:///mnt/bigdata/tpch/sf100",
+    "queries": {
+        "2": 8.547899007797241
+    }
+}

tpch/datafusion-ray-tpch-1739638680909.json

@@ -0,0 +1,35 @@
+{
+    "engine": "datafusion-ray",
+    "benchmark": "tpch",
+    "settings": {
+        "concurrency": 8,
+        "batch_size": 8182,
+        "prefetch_buffer_size": 0,
+        "partitions_per_worker": null
+    },
+    "data_path": "file:///mnt/bigdata/tpch/sf100",
+    "queries": {
+        "1": 7.222118854522705,
+        "2": 8.797776937484741,
+        "3": 11.183124780654907,
+        "4": 7.6282007694244385,
+        "5": 20.619840383529663,
+        "6": 3.466888427734375,
+        "7": 29.999598026275635,
+        "8": 22.716665267944336,
+        "9": 38.37256050109863,
+        "10": 25.540525197982788,
+        "11": 6.4380128383636475,
+        "12": 10.021047592163086,
+        "13": 8.462335348129272,
+        "14": 3.810248851776123,
+        "15": 0.8309383392333984,
+        "16": 3.692992925643921,
+        "17": 32.96640610694885,
+        "18": 50.401840925216675,
+        "19": 4.988840818405151,
+        "20": 7.992424011230469,
+        "21": 47.60438632965088,
+        "22": 3.4463324546813965
+    }
+}