We describe IoTCloud, a platform to connect smart devices to cloud services for real time data processing and control. A device connected to IoTCloud can communicate with real time data analysis frameworks deployed in the cloud via messaging. The platform design is scalable in connecting devices as well as transferring and processing data. With IoTCloud, a user can develop real time data processing algorithms in an abstract framework without concern for the underlying details of how the data is distributed and transferred. For this platform, we primarily consider real time robotics applications such as autonomous robot navigation, where there are strict requirements on processing latency and demand for scalable processing. To demonstrate the effectiveness of the system, a robotic application is developed on top of the framework. The system and the robotics application characteristics are measured to show that data processing in central servers is feasible for real time sensor applications.
2 July 2017 in The International Journal of High Performance Computing Applications
With the ever-increasing need to analyze large amounts of data to get useful insights, it is essential to develop complex parallel machine learning algorithms that can scale with data and number of parallel processes. These algorithms need to run on large data sets as well as they need to be executed with minimal time in order to extract useful information in a time-constrained environment. Message passing interface (MPI) is a widely used model for developing such algorithms in high-performance computing paradigm, while Apache Spark and Apache Flink are emerging as big data platforms for large-scale parallel machine learning. Even though these big data frameworks are designed differently, they follow the data flow model for execution and user APIs. Data flow model offers fundamentally different capabilities than the MPI execution model, but the same type of parallelism can be used in applications developed in both models. This article presents three distinct machine learning algorithms implemented in MPI, Spark, and Flink and compares their performance and identifies strengths and weaknesses in each platform.
30 June 2016 in International Journal of Grid and Distributed Computing