Praveen Pavithran

It makes no sense if you use MongoDB primarily as a relational database. As you scale MongoDB will be more expensive than SQL and as you said without having the advantages of "join" etc.

We use MongoDB in our company. It is useful for us, as we work with different types of devices and we love the functionality of being able to add fields whenever we have a new device type etc. Mongo also allows enables easy scaling and fault tolerance. However, you will have to learn how to manage it.

If you are already comfortable with SQL and don't need NoSQL, stick to SQL. At scale, it is cheaper than Mongo.

The DB and some servers on a separate sub-net in the VPC. This ensures access to these servers are denied from any other machine than the VPC. Amazon VPC

All images are stored on S3. These images include map images of trips, and graphs of data analysis. Amazon S3

Fast cache is used to store short term information from devices and detect change in status of the device. Redis

Unstructured No-SQL DB was important as deal with several different devices. MongoDB

Maps APIs for addresses, making image summaries of trips, view vehicles on a various front-ends. Google Maps

We use several EC2 instances each dedicated for a separate purpose - API servers, TCP servers for device, DB instances. Amazon EC2