Rocket Ties Analytic Database Into IBM i
June 13, 2016 Alex Woodie
IBM i shops that are interested in extracting insights from their data under the guidance of automated statistical analysis and machine learning may want to check out a new offering from Rocket Software. Called Rocket Discover for IBM i, the new software combines data discovery capabilities with self-service data preparation, and is aimed at regular business users who find things like Tableau Software‘s products too geeky or difficult to learn.
Rocket began building its Rocket Discover tool several years ago to serve the growing need for self-service tools that can handle the entire analytic spectrum–from data preparation and discovery to data visualization and collaboration.
The idea was to make the software easy enough to use without requiring help from the IT department or extensive classroom training, but also powerful enough to enable users to drill-down from aggregate-level views and go wherever the data takes them.
“We really wanted to target business users,” says Patrick Spedding, managing director of BI R&D for Rocket Software, and a 20-year veteran of the BI industry. “A lot of the tools on the market target the data geek, the business analyst.”
Spedding used to sell Tableau’s software in a previous job, and while he appreciates their power, he also understands they’re not for the faint of analytic heart. “You have to go to a training course to really make the most out of it,” he tells IT Jungle. “It’s designed for somebody who is a data scientist or somebody who really understands data well.”
Rocket Discover for IBM i
Rocket Discover for IBM i includes two primary components. On the back-end is an in-memory columnar database that runs on Windows and Linux. The database is built in part using open-source components, including a Redis key-value store and a MongoDB NoSQL database, to store the data.
On the front-end is a Web-based user interface developed in HTML5 and AngularJS. Whereas other BI/data visualization tools use full Windows clients, Rocket decided to use the latest Web-based technologies to help make the product as simple and easy to use as possible. “We’re trying to offer functionality in a more beautiful, simplistic way,” Spedding says. “We focused a lot on simplicity. Imagine Steve Jobs as the end-user consumer. It has to be that beautiful thing that he would want to use.”
Rocket is aiming for a slightly less sophisticated user with Rocket Discover. That includes the newly announced Rocket Discovery for IBM i, which adds the DB2 for i database connector that Rocket announced at the recent COMMON conference in New Orleans, Louisiana.
As it did for other pieces of the suite, Rocket develops its own database connectors for IBM i, System z, and the MultiValue database (Rocket acquired the MultiValue database from IBM in 2013). It also sports connectors to other analytic databases, such as Teradata, Netezza, Oracle, and Microsoft SQL Server. While building custom database drivers takes time and effort, it’s worth it to avoid generic ODBC drivers at the end of the day, Spedding says.
“It was potentially going to be a benefit to build something that was actually optimized for those environments,” he says. “If you go with a generic ODBC tool, you’re going to end up losing part of the reason that somebody bought a particular application. You’re losing the understanding of the hierarchies or the rollup rules or whatever it happens to be. If you’re flattening that, you’re losing some of the application value.”
Once the data is pulled into Rocket Discover, the tool helps users by guiding them through various types of analysis. Once again, Rocket has focused on flattening the learning curve without dumbing down the end product.
“We use descriptive statistics to try and uncover patterns. But we’re not trying to be a full blown Watson type tool,” Spedding says. “We started off with simple outlier detection. The user doesn’t have to know how to write an R function or call an R script in order to be able to do NV outlier or whatever it is. We said, let’s just make it a check box in the tool.”
Raw data is typically not clean enough for high-level analysis, so the first thing that every user has to do is prepare the data. To that end, Rocket Discover includes several self-service data preparation capabilities that can make this data “janitorial work” go faster.
“When we read data in, we can do automatic hierarchy detection to suggest natural hierarchies in the data,” Spedding says. “We have some correlations we built in so if you want to compare a couple measures, it will help you see if there’s a correlation between those.”
When the data prep is done, Rocket Discover can provide more powerful capabilities, such as being able to overlay clustering on a scatter-plot to detect at-risk customers who are likely to churn using a basic k-means algorithm. “Instead of having a user have to figure out how to use it, we can actually bake it into the user interface,” Spedding says.
“We kind of analyze each piece,” he continues. “We’re trying to do it in such a way that we’re not throwing the entire gamut of analytic on a poor business user, but we’re trying to make it something that’s usable and useful.”
In addition to serving dashboards, Rocket includes the capability to share insights. That’s where its integration with IBM Connections comes in.
“Collaborative BI is the next wave,” Spedding says. “The whole point of BI is to share your insights. If you just build a dashboard and email it to someone as a PDF, you’re not really doing service to that analysis. Where we’re really focused is pulling a dashboard into a chat session and having a bi-directional conversation about it and actually incorporate it, through collaboration, with data discovery and data governance. Those are our big themes at the moment.”
Rocket Discovery for IBM i is available now. There are several licensing schemes, including named-user pricing starting at $1,000 per user per year. Unlimited user pricing is available for $12,500 per core. For more information see www.rocketsoftware.com.