CIRUS : A Cloud Infrastructure for Real-time Ubilytics (ubiquitous big data analytics)
In this talk, we will present the CIRUS, a platform to build domain-specific PaaS platforms for real-time Ubilytics. We will demonstrate CIRUS in the context of two Internet of Things (IoT) application domains: SmartGrid and Smart Cities.
The Internet of Things (IoT) has become a reality with the availability of chatty embedded devices (RFID, wireless sensors, mobile sensors, personal smartphones…). It provides to companies new opportunities of economic models (i.e., pay as you use), improves the quality of service delivered to their customers (individuals or companies), and helps them to satisfy their legal and contractual duties.
The IoT is now an established efficiency and productivity tool for agile organizations. However, associated services called Machine to Machine (M2M) require the analysis of a huge amount of data produced by swarms of sensors and collected by dedicated gateways. This huge amount of IoT data must be analyzed with models (Map-Reduce, ESP, CEP) and technologies (Hadoop, Storm, …) of the “Big Data Analytics”, deployed on Cloud platforms. This new field is named ubilytics (ubiquitous big data analytics). Configuring and deploying the infrastructure for ubilytics is a complex task since several skills are required to deal with various technologies (from embedded M2M gateway to cloud-based analytic platform).
The CIRUS project aims to build domain-specific PaaS platforms for real-time Ubilytics.
The CIRUS infrastructure collects and analyzes IoT data from M2M gateways to cloud hosting for agile design and deployment of M2M services.
This platform should
* alleviate the ubilylists (business experts who analyze ubiquitous big data) work in the configuration of the components involved in the collection and filtering infrastructure from the M2M gateways (such as OpenHAB, OM2M, IoTSyS, …) to the streaming processors deployed on a cloud platform.
* provide self-adaption and elastic mechanisms in order to support variations in the sensors data throughput and to supports virtual machine faults.
* provide benchmark tools in order to stress the infrastructure before the production stage.
* provide statistical/numerical libraries for streaming processors topologies fitting domain specific concerns (ie smartgrid consumption forecast, …)
* provide DSL-based plugins for Eclipse in order to design easily the Ubilytics queries (sensors to use, topologies to reuse, protocols to use, …).
The two demonstrations of CIRUS involve both OpenHAB, Mosquitto, RabbitMQ and Storm. The OpenHAB home management platform is running on Intel Galileo embedded boards for collecting and filtering sensor data (electricity consumption, temperature, presence, ...). The sensor data are sent using the MQTT protocol through the MQTT brokers (Mosquitto and RabbitMQ). The Apache Storm is hosted on the Windows Azure cloud platform for the realtime analytics. These components are deployed by Roboconf (an open-source cloud deployment manager developed by LIGLab and Linagora http://www.roboconf.net ) on virtual machines hosted by the Azure IaaS and onto the Intel Galileo embedded gateways which can be installed in real houses, stores, offices, buildings and cities.