Perform TPC-DS Benchmark Testing

This guide describes how to run the TPC-DS benchmark on SynxDB Elastic v4.x.

Prerequisites

Ensure the following conditions are met before running the test from a remote client host:

psql client is installed and passwordless access to the remote cluster is configured (via .pgpass).
The gpadmin database is created with the default warehouse parameter:
```
ALTER ROLE gpadmin SET warehouse = 'testforcloud';
```
Replace gpadmin and testforcloud with the real user and warehouse you are using.

Step 1. Install TPC-DS tools dependencies

Install the required dependencies on mdw:

ssh root@mdw
yum -y install gcc make byacc

The original source code is available at: http://tpc.org/tpc_documents_current_versions/current_specifications5.asp.

Step 2. Download and install TPC-DS tools

Clone the TPC-DS tools repository from GitHub:

ssh gpadmin@mdw
git clone https://github.com/SynxDBFE/TPC-DS-CBDB.git

Step 3. Get psql connection options

To run the benchmark, get the necessary connection options for PostgreSQL (psql).

Log in to the Kubernetes (K8s) cluster and identify the tenant namespace:

kubectl get ns

Example output:

NAME                                           STATUS   AGE
default                                        Active   97d
synxdb-system-4x                               Active   96d
kube-node-lease                                Active   97d
kube-public                                    Active   97d
kube-system                                    Active   97d
storage-system                                 Active   97d
tenant2-313a164c-cfe9-4514-941d-4c1d548ddf97   Active   93d

Obtain the cluster mapping ports for the tenant:

kubectl get svc -o wide -n tenant2-313a164c-cfe9-4514-941d-4c1d548ddf97

Example output:

NAME                    TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE   SELECTOR
cloudberry-proxy        ClusterIP   None           <none>        5432/TCP         93d   app=cloudberry-proxy
cloudberry-proxy-node   NodePort    10.96.199.21   <none>        5432:32355/TCP   93d   app=cloudberry-proxy

In this example, the obtained port is 32355, and the IP can be any node in the K8s cluster (for example, 10.14.3.242).

Step 4. Modify configuration files

Edit the tpcds_variables.sh file to configure the test parameters. Fill the PSQL_OPTIONS with the real psql connection information you have obtained. Edit GEN_DATA_SCALE according to your cluster size and need.

ssh gpadmin@mdw
cd ~/TPC-DS-CBDB
vim tpcds_variables.sh

Example configuration:

export RUN_MODEL="cloud"
export RANDOM_DISTRIBUTION="true"
export TABLE_STORAGE_OPTIONS="compresstype=zstd, compresslevel=5"
export CLIENT_GEN_PATH="/tmp/dsbenchmark"
export PSQL_OPTIONS="-h 10.14.3.242 -p 32355"
export GEN_DATA_SCALE="10"  # Generate 10GB of data.
export MULTI_USER_COUNT="5"   # Number of concurrent streams for the throughput run.
export RUN_SINGLE_USER_REPORTS="true"
export RUN_MULTI_USER="true"
export RUN_MULTI_USER_QGEN="true"
export RUN_MULTI_USER_REPORTS="true"
export RUN_SCORE="true"

For additional configuration parameters, refer to the repository README file: https://github.com/SynxDBFE/TPC-DS-CBDB/blob/main/README.md.

Step 5. Run the TPC-DS Benchmark

To execute the benchmark, log in as gpadmin and run the test script:

ssh gpadmin@mdw
cd ~/TPC-DS-CBDB
./run.sh