In this blog, experts from Fuzzy Logix & Hortonworks describe the challenges the Manufacturing industry is faced with, use cases and benefits of analyzing data in place with solutions from Fuzzy Logix and Hortonworks.
Manufacturers are on the forefront of re-imagining how Big Data can redefine industry economics. In recent years manufacturing Big Data volumes have exploded, arising from sensors on machines and assets, in addition to real-time data from logistics, sales and marketing systems. This data can amount to hundreds of terabytes of data that can transform manufacturing and supply chain operations.
While Big Data provides great benefits, it can also introduce significant challenges. For example, when performing analysis on these huge data sets, traditional reporting and BI solutions often require moving terabytes of information from where it was stored to the analytics application itself. This creates significant challenges. First, large volumes of data must be moved across precious network resources and, perhaps more challenging, enterprises must now store and manage duplicate sets data, requiring considerable additional security and administrative processes to manage two separate environments.
However, what if you could perform your analysis where the data is located, and not move it all? The good news is that this is possible – using proven technology from Hortonworks and Fuzzy Logix.
Hortonworks open data platforms allow companies to store massive amounts of data. As the industry leader in providing solutions that blend the best of both open source and partner applications, Hortonworks provides state of the art capabilities to manage the data associated with real-world Big Data initiatives. In addition, as a certified Hortonworks partner, Fuzzy Logix provides a library of hundreds of parallelized high-performance algorithms that simply installed into existing Hadoop nodes. In effect, our software becomes part of the operating system; just as native functions work within traditional databases. To access the models, you simply write SQL and viola – you are analysing your data on-demand – without moving it and with no additional hardware, network or infrastructure cost.
What happens when you analyse data without moving it? A few key benefits emerge:
- You no longer need wait for data to move to perform analysis and all of your analysis will be on the most current data.
- Time to perform analytics is typically 10X or 100X faster.
- You will be able to analyse data at an almost unlimited scale.
- Data scientists can build models and deploy them to end users by embedding the models into existing reports or BI tools.
- You will have access to hundreds of certified machine learning, data mining, text, math and statistical models that have been developed and market-proven for over 10 years.
What kinds of things are companies achieving using high-performance Fuzzy Logix models running on Hortonworks solutions?
- Demand Analytics: A large retailer needed to minimize waste in their perishable foods. Prior to working with in-database analytics, they could only perform analysis at regional and category levels over a period of days. After deploying in-database analytics, they were able to perform store and SKU level analysis within hours. The analysis leveraged widely ranging data, including weather, to forecast demand and spanned thousands of stores and thousands of SKUs. To generate these results, over 13 million models were running simultaneously. As a result, the retailer was able to increase product availability by over 4 percent, resulting in millions of dollars in revenue gain per year.
- Predictive Maintenance: A medical device manufacturer wanted to build an analytics platform for the future to support predictive maintenance for thousands of machines and devices. The goal was to predict the optimal run time of a device or component before failure, balancing the need to both prevent failures of these parts while also maximizing the service life of these parts, reducing waste. Utilizing the “in-Hadoop” analytics approach, the manufacturer achieved a 100x analytics performance improvement, providing the ability to model very large data sets at a very granular level of detail, providing key insights that resulted in huge cost savings.
- Quality/Yield Optimization: A chip manufacturer wanted to improve manufacturing yield. Consider the fact that a 1 percent improvement in yield is worth approximately $100 million in the semiconductor business. The machines used in semiconductor manufacturing generate huge amounts of data that can be analyzed to reduce defects and optimize manufacturing yields. Using in-Hadoop models, the manufacturer was able to transform from analyzing sample data to working with full data sets, providing the ability to find problems that were invisible when constrained to small data sets.
The above case studies represent just a few examples of how analytics can be leveraged to reduce manufacturing costs and drive revenue optimization processes. Should you be interested in investigating solutions to challenges occurring within your company, please contact us. We would be happy to discuss your issues in detail or conduct a ½ day analytics workshop to help you understand how to get the best value from running Fuzzy Logix analytics on Hortonworks. We can also provide our software to you for testing, providing you the ability to see results for yourself, using your very own systems and data.