How to repeat the tests

We have prepared several AWS AMIs for anyone who is interested in replication of these tests. These AMIs contain some tuned OS parameters, for example, the maximum amount of threads per process and open file descriptors:

If you would like to verify performance for all components hosted on a single server instance, simply run NDU AMI instance. By default this instance will be using Cassandra that runs locally.

If you would like to verify performance for standalone NDU server that uses external Cassandra Cluster, please init Cassandra cluster using provided Cassandra AMI first. For example, let’s do the configuration for three Cassandra instances. Once you have launched 3 AWS instances using Cassandra AMI please update cassandra.yml file to make them run in a cluster. In our case we have 3 instances with the following IP addresses:

Login into every cluster instance, clean up cassandra data directories and modify cassandra configuration:

1
2
3
4
sudo rm -rf /var/lib/cassandra/saved_caches/*
sudo rm -rf /var/lib/cassandra/commitlog/*
sudo rm -rf /var/lib/cassandra/data/*
sudo nano /etc/cassandra/cassandra.yaml

Find in the file next lines and update them accordingly.

For instance A:

1
2
3
4
5
seeds: "172.21.12.100,172.21.12.101,172.21.12.102"

listen_address: "172.21.12.100"

rpc_address: "172.21.12.100"

For instance B:

1
2
3
4
5
seeds: "172.21.12.100,172.21.12.101,172.21.12.102"

listen_address: "172.21.12.101"

rpc_address: "172.21.12.101"

For instance C:

1
2
3
4
5
6
seeds: "172.21.12.100,172.21.12.101,172.21.12.102"

listen_address: "172.21.12.102"

rpc_address: "172.21.12.102"

On every instance restart cassandra:

1
2
sudo service cassandra stop
sudo service cassandra start

And verify that Cassandra cluster setup was successful:

1
nodetool status

In the output should be something similar:

1
2
3
4
5
6
7
8
Datacenter: datacenter1
=======================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address        Load       Tokens       Owns (effective)  Host ID                               Rack
UN  172.31.28.47   192.99 KiB  256          30.9%             a323e6fb-2e8c-4bb4-82d0-4e621cb7cba8  rack1
UN  172.31.19.231  132.23 KiB  256          33.9%             6da17a19-2a4b-4f99-9ac7-e38f05ebf7a9  rack1
UN  172.31.25.178  289.4 KiB  256          35.2%             87f1ab4d-16d4-4969-aea8-b858e62d1d73  rack1

Once the cluster is ready we need to create schema, system and demo data. At any of the Cassandra cluster node (here we use instance A) execute following commands:

1
2
3
cqlsh 172.21.12.100 -f /usr/share/thingsboard/data/schema.cql 
cqlsh 172.21.12.100 -f /usr/share/thingsboard/data/system-data.cql 
cqlsh 172.21.12.100 -f /usr/share/thingsboard/data/demo-data.cql 

Once Cassandra cluster setup is done please run NDU AMI instance. You need to update thingsbaord.yml config to use Cassandra cluster instead of a local instance:

1
sudo nano /etc/thingsboard/conf/thingsboard.yml

And update the cassandra url from localhost to IPs of cassandra ring:

1
url: "${CASSANDRA_URL:172.21.12.100:9042,172.21.12.101:9042,172.21.12.102:9042}"

After configuration update, restart NDU service:

1
2
sudo service thingsboard stop
sudo service thingsboard start

Once you will setup your cluster configuration using NDU and Cassandra AMIs you can execute tests from “client” machines (NDU Performance Test AMIs) using following commands:

1
cd projects/performance-tests

Update mqttUrls and restUrl and set private IPs of AWS instance where NDU service is deployed in test.properties file:

1
nano src/main/resources/test.properties

Re-install project so Gatling can pick up latest config files and start test:

1
mvn clean install gatling:execute