Today, most of the “big data” applications needs to compute data in real-time since the Internet develops quite fast and the users expect the get reactions from the applications simultaneously. This rule is valid for almost all types of applications. When a user interacts with a commercial website by looking a product, the website should be able to show her related products for increasing its conversion rates. For a CRM application, the users should be able to solve their problem using that application. And most the time, these actions need aggregation computations. The aim of our project is to provide a high-performance scalable computational engine that is flexible and can be adapted to any type of applications. The system collects the events with collection API and continuously processes them on the fly with using pre-aggregation rules submitted by Analysis API.
By adapting a previously written percolation model in C, the threshold probabilities for square, triangular, and cubic lattice types were confirmed. An algorithm to count the distribution of cluster sizes at a variety of percolation probabilities was developed, and the expected trends towards the so called infinite cluster was achieved. An equivalent bond percolation model was adapted to the original site algorithm, and by treating occupied bonds as springs, a total compression trend for the model was constructed, which implied that structures under the boundary conditions that were imposed does not have behavior that changes the total compression constant significantly at the percolation threshold.\\