DeploymentAndConfiguration

#Deployment and Configuration

Obviously, Dempsy is meant to be deployed in a distributed manner. It's designed to make this really easy, yet not take on responsibilities better solved by other infrastructure components. By this we mean that Dempsy does NOT operate as an "Application Server." You do not "deploy" an application to Dempsy. Rather you deploy your application, along with Dempsy.

There is a significant difference between batch oriented big-data systems and stream based big-data systems. In batch oriented big-data (Hadoop) you have data partitioned among nodes and you send the functionality to the data in order to operate on it in parallel. Therefore, in cases like that, the concept of "deploying" and application to an application server makes sense.

Here, however, your functions run, and the job of the system is to get the data to your code. This means there can be a better separation of concerns and well established IAAS cloud based provisioning tools can be used to manage the system.

Dempsy is designed to be run in a cloud based system and rely on the provisioning tools already supporting the cloud. On Amazon, for example, you can automatically provision more nodes in response to load (using Amazon's "CloudWatch").

Dempsy is responsible for the coordination of nodes, once they are started. Not for starting them. But Dempsy makes this coordination work with almost no effort or configuration. The next sections describe how this works.

Next section: Distributed Deployment