AWS best practices for allocating VMs to processes

What is the best practice to allocate EC2 VMs for a multi-tier web application.
I have the following processes

• Node.js app for web-server
• Apache Kafka and Storm for backend. (or consider any middleware framework for that matter)
• Elasticsearch for storage
• Zookeeper for co-ordination

Should I put each of the above in its own VM so that its easy to scale and maintain? Is that how it should be done?
What other approaches do people follow so I can make efficient use of the VMs?

Thank you