-
-
Notifications
You must be signed in to change notification settings - Fork 604
Benchmarking Datastax Cassandra cluster with YCSB on EC2
Tzach Livyatan edited this page Jul 17, 2014
·
8 revisions
Closely following the Datastax instructions
I manage to run, but not benchmark, a C* cluster. Here is how I did it, what didn't work, an important notes on Cloud-init and a existential question.
- Launch AMI ami-ada2b6c4(HVM) or ami-f9a2b690 (PV) For large deployment, use PV (see below)
- Use cassandra-cluster security group
- Choose one of the recommended instance:
- Development and light production: m3.large
- Moderate production: m3.xlarge
- SSD production with light data: c3.2xlarge
- Largest heavy production: m3.2xlarge (PV) or i2.2xlarge (HVM)
4.Choose number of VM to run, and set user data (3 in the following example)
User data
--clustername name test-cluster
--totalnodes 3
--version community
--opscenter no
- Add Storage
- Launch
You now have a running Cassandra cluster
The nodes find each other automatically using a centralize Datastax service.
To validate cluster status, ssh into one of the cluster members and run
nodetool status
This will present a list of the cluster members:
Datacenter: us-east
===================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
-- Address Load Tokens Owns (effective) Host ID Rack
UN 10.233.78.197 40.97 KB 256 71.1% edae6d05-4f12-4d57-b518-9b83d1bee493 1c
UN 10.79.142.68 40.94 KB 256 64.0% aa30ea76-5ec7-41c4-b457-3e88bd6904d9 1c
UN 10.146.227.234 40.95 KB 256 65.0% ec0fccbb-63b7-42ce-9f8b-17ea321b830c 1c
- Create a EC2 Linux instance
- follow the instructions to test the cluster. Make sure to use
strategy_options = {replication_factor:3};
to take advantage of the cluster replication replication factor of 3 is used by Google C* benchmark