IoT is a computational notion where each of the tangible objects around are termed as things, which will be connected to other things through the web and will communicate or transfer data between each other. But if the focus is just on the things or devices, than there may be a key ingredient missing, that is all these internet of things are connected to each other. What this entitled is that all of the devices, from appliances to smart phones will be able to share information with one another. When gone into details about Internet of things (IoT), it has lots of different types of things it has devices, it has locations, it has people and they all are connected. That mean a big and a complex connection data is generated. When to model this connected data a relational database or other type of No SQL database is used, they under perform. As those relational and No SQL type abstractions are not optimized for connection data, they are optimized for different kinds of connections. To overcome this, a highly robust, scalable, pluggable and faster architecture is proposed which tries to solve the issues of connected data in IoT domain with the unification of three open source technologies such as Ejabberd, Apache spark and Neo4j database. And to a get more secure and efficient performance this three open source technologies are implemented on a Microsoft azure public cloud. This proposed architecture stack is equally suitable for projects with high ambitions.