Monday, October 24, 2016

Robo4j introduction to the reactive "real-time" micro-services based IoT system development

Robo4j testing system : camera unit, ultra-sonic, gyro, color and touch sensors. Whole system contains also one  Raspberry Pi, Lego Brick and CuBox i4Pro (4-core CPU) with 500GB hard-drive as re-usable data-storage. Whole system is powered by 2x2500mAh and one 25000mAh unit.
  What is going to be discussed in this article is the high level perspective of the upcoming framework and the Internet Of Things (IoT) system design, so let’s move forward.
  Over the last few years has been one of the most emerging topic micro-services (and still is) followed by reactive manifesto, which explains how to design non-blocking highly responsive, resilient, elastic and message driven parts of the application. From my perspective, all those articles have been focused on the web applications of middle and big sizes. Most of the articles have been describing the best practices about how to split the monolithic systems or how to employ the news trends in micro-services and non-blocking design patterns. 
  Only few of them have been touching other emerging technologies around Internet of Things (IoT) in the way I’ve been searching for. 
  Over the years, reading many of them, I got a strong feeling that the way how we intend to develop IoT systems is, that we take already developed robust technologies and we make them smaller.  These days we have on the market good examples of the successes, we do use Spring-boot, Netty, Vert.x or others.  Those frameworks allow us basically to connect the developed system with different types of databases, analysed and react on events or data, obtained from the data storages. 
The question is still there,  is this really enough ? 
Is such kind of abstraction good enough that it could allows us to develop robust IoT systems ? 
  The IoT systems that can react on incoming events, sorts them properly and in special cases forwards them to the proper system parts, which could process and execute such events ? 
  I don’t think so or I’ve not found anything like this on the market.

  We can go further in questioning ourselves about the IoT possibilities, which can quite soon turn into the Artificial Intelligence discussion. In my opinion, the AI question is not relevant for the article, because even most of the currently available IoT systems can’t still exhibit its intelligence in the manner of intelligent (independent) decisions. This question is reserved for another IoT focused article. 
  Let’s go back to the main topic: advanced Internet of Things system development and design. Before we start here a small review of the definition taken from Wikipedia: 
   “The Internet of Things (IoT) is the internetworking of physical devices, vehicles, buildings and other items—embedded with electronics, software, sensors, actuators, and network connectivity that enable these objects to collect and exchange data”. In other words IoT are objects that are connected to the internet or other networks and those objects are allowed to receive and process data. 

   The definition looks pretty cool and good so far, until you try to use some of those connected “smart” objects together. Now is the IoT system the collection of the independent IoT units and suddenly nothing works as expected due to connection instability, communicaiton protocols message ordering etc. 
  Although such a system has been assembled from cutting edge parts, altogether they simply don’t work, due to previously touched issues or other ones we are not even aware of. The whole development may turn into the really FUZZY mode.  
Such a style of development I normally call try, push, deploy development (TPDD) but this style may not fully satisfy the expectations or needs, it simply doesn’t work in a long term perspective. PS: there is still a chance that I may be wrong.
The described situation can become really frustrating and stressful, especially when there are not so many possibilities on the market. You may use ROS (Robot Operating System) but … let’s go back.

  The lack of proper tools on the market was main motivation to develop framework which satisfies the following needs: 
  1. The is capable to synchronise incoming asynchronous events/messages under the dynamically changing conditions at “real-time” . 
  2. The framework should be able to allow and enforce rapid micro-sevices based system development.
  3. The should be a light framework that is capable to run on any hardware supported by JVM. It means on any Java Powered device and it provides full performance
  4. The should be independent, not hardware specific, it must be flexible to extend
  5. The should have a control over all running units inside the eco-system 
  6. The should enforce communication with external systems
  7. The should be easy to use
   All those seven mentioned points are really challenging. Any of them is trivial as it may seem.  I’ve taken this challenge and I’ve developed such a tool that satisfies all of them. I’ve developed already the mentioned framework.
  The framework allows to connect and control any device that is accessible over the specific type of network (Ethernet, WiFi, Bluetooth etc.) or wire. And moreover, enforce the usage of any of technology running on JVM with its natural language (Scala, Groovy etc.).
  Now it’s time to take a look at the basic framework architecture from a high level perspective which is the main intent of this article. The image below shows the high perspective reaction between the different parts inside the robo4j eco-system. It’s important to see the differences between the inputs to the system.
Robo4j framework high-perspective architecture
  Such input could be provided by Sensors, Motors (Engines) or RoboSockets types. The incoming messages is then processed by the system itself and serialised in the manner that the result can be executed on the connected hardware.
Inside the message process stage the system may use different types of datastorages that are available to a specific RoboUnit. This means, that each RoboUnit may have or not it’s specific data resource or each RoboUnit may share them with another one in reactive, non-blocking manner. 
  The Robo4j Message Bus is not dependent on the order, how any specific message is processed. The bus consumes the messages in order of their arrival. The bus may dynamically change the conditions in real time. The pre-results are then moved into the MessageReactors parts. Those parts have again connections to the datastorages, where their results can be stored. The final result can be executed on the real hardware or can involve the whole system state inside the predefined cloud. 
  Finally we are at the end of the motivaiton article about framework. The whole text should motivate you to use it, because the Internet of Things is the future of IT. 
The future of IoT includes the communicaiton security, different kinds of messages processing and reaction on them. The Robo4j framework is supposed to be part of this future.
Stay tuned.
Robo4j IoT system (robot) is discovering a park


No comments: