Karmasphere has an interesting story to tell. Much like Datameer, which I recently blogged about, Karmasphere sits on top of the Hadoop distributed platform where companies such as Cloudera, Hortonworks and MapR compete. Karmasphere provides a collaborative environment and an analytical workbench that help companies write applications and workflows that run on top of Hadoop. The company’s business model looks to leverage legacy skill sets, such as SQL, which are already resident in most organizations, in order to ingest, analyze and act on big data.
Karmasphere’s approach begins with the common assertion that business intelligence tools were built to analyze only structured data. They use descriptive statistics and provide historical views of data, but they are limited in their iterative discovery processes and in their ability to add new data in a timely, practical manner. Newer tools, such as in-memory databases and appliances, address the old technological limitations but have their own issues. While memory is becoming less expensive and speed is improved, proprietary hardware can lock businesses into a particular technology, and the way these tools manipulate data is often proprietary, too, in terms of how data is written back to disk and what types of data are stored on disk.
It’s through this gap in the market that Hadoop developers and companies like Karmasphere want to capitalize. Hadoop provides an ideal platform for companies exploring and analyzing big data because it is built to maximize disk I/O, run on commodity hardware and scale in a linear fashion. Given that commodity hardware by definition is fast, cheap and available, Hadoop clusters fit the bill for advancing big-data analytics.
The market is responding to this logic. Not only are we seeing venture capital pouring into the space, but our own benchmark research on Hadoop shows that it is being used in 33 percent of companies’ big-data environments and evaluated in another 20 percent. Very likely these numbers are increasing as I write this.
A big issue for Hadoop adoption is the skills gap. In our research on business analytics, 89 percent of participants said it is important to make it simpler to provide analytics and metrics to all users who need them. Last year big data was the domain of ninja data scientists creating MapReduce functions. Now, while big data is still a nebulous concept for many business people outside the data world, it is starting to come into focus, and the conversation is moving to the front lines of organizations. The discussion now is about how to match the power of Hadoop with the current skills of developers, data analysts, business analysts and end users.
Karmasphere addresses the skills issue. The company previewed its 2.0 release at the Hadoop Summit in June. The release focuses on the collaborative workspace and includes a shared repository for analytical assets in a role-based Web environment. With this team-oriented focus, Karmasphere is hitting a sweet spot in the market, as end users are becoming more involved in both usage and buying decisions. We’re looking forward to discussing this trend in our next-generation business intelligence benchmark research, due out soon.
As you might expect with an integrated Hadoop approach, Karmasphere 2.0 can ingest data from multiple sources, including data from Omniture and DoubleClick. Its wizard-based approach is helpful in identifying and consuming diverse data sources into the Hive tables native to Hadoop. The software enables users to do basic descriptive data exploration to determine the techniques and algorithms that they might want to apply. From there, analysts can conduct ad-hoc queries and iterative SQL analyses to test hypotheses and find insights from the data. Karmasphere also includes support for Hive UDFs and analytics packages such as SPSS and SAS for more complex analysis. Finally, Karmasphere helps users embed their insights in other systems through REST APIs and to publish through other BI applications such as Tableau and Spotfire.
By positioning the software more toward collaboration and analytics and less toward visualization, Karmasphere appears to be trying to carve out a unique space and avoid direct comparisons with companies such as Datameer that have advanced analytics and visualization capabilities. Karmasphere’s approach to visualization seems to be to provide some basics and partner with firms such as Tableau to provide more robust front-end user tools. This strategy allows Karmasphere to focus both its messaging and development efforts on the core value of bridging the Hadoop skills gap with analytics and collaboration tools.
Karmasphere has a strong story, but broader questions remain about how the Hadoop ecosystem will evolve from early adopters to early majority. A lot of money is still flowing into the Hadoop community, but I anticipate a shakeout in the space at some point, and only vendors that provide strong time-to-value propositions will remain intact. Karmasphere provides a compelling proposition around team analytic processes, collaboration and bringing the value of Hadoop to the front lines of organizations. By competing to fill the space between the technical aspects of Hadoop and the existing skills in the market, it manages to address a critical challenge around adoption.