Can Isima Be the Nutanix of Data Management?
Despite the technological advances in big data, companies continue to struggle to put it all together and manage data in an effective way. Now a company called Isima is stepping forward with a plan to build a platform that combines multiple data management disciplines–including ESBs, EDWs, ETL, and BI–into one hyperconverged system, or what its CEO dubs “the Nutanix of data management.”
We’ve gone through three distinct architectural phases in building big data systems, according to Darshan Rawal, the CEO and co-founder of Isima.
The first phase, which started in the late 1990s, was based on a scale-up client-server architecture that utilizes master-slave relationships and standard relational databases. Oracle, Microsoft, IBM, and the rest built enterprise IT solutions atop this architecture, and many of them are still in place today.
The second phase began with the publishing of the Google Big Table in 2006, which spurred numerous distributed systems like Apache Hadoop, Apache Cassandra, and other “big data” technologies. These scale-out systems, such as Cloudera’s distribution of Hadoop, could store petabytes of data, but were hampered by the co-location of compute and storage (although Cloudera, of course, is moving past Hadoop).
We’re currently in the third phase, which is marked by the rise of cloud-native architectures from the cloud giants that can scale compute and storage separately, as well as advances like microservices and containers. Systems like Snowflake and Google Big Query are good examples of this style of architecture.
But despite the progress, companies are still largely using the same sorts of tools to move and manage data, Rawal says. “Data management for these three generations have not be consolidated,” he says, “and we believe that systems innovation will allow all of these things to come together.”
This is the basis for Isima’s offering, called bi(OS), which aims to radically streamline and simplify the software stack that companies today use to store and process data. By consolidating the four broad category of tools–enterprise data warehouse (EDW); extract, transform, and load (ETL); enterprise service bus (ESB); and business intelligence (BI)–Rawal hopes to provide a much shorter path to downstream data innovations, including AI.
“We look at the entire stack, from ingest to insight, how can we deliver it with one engineer. You don’t need think about the individual pieces across the whole stack,” he says. “That’s what Isima really does, and that’s why we feel like we are the Nutanix for data management.”
Nutanix, of course, delivered a hyperconverged infrastructure platform that enables customers to get the benefits of cloud-scale but without all the hassles and overhead that go along with running in the cloud. “Nutatix did exactly this in 2008,” Rawal says. “They went against the cloud. They said, enterprises don’t need scale. All they really need is a hyperconverged infrastructure platform, and we are doing the same for data management.”
Of course, companies are not going to rip out 30 years’ worth of data infrastructure and replace it with a relatively new system, such as bi(OS), based on a PowerPoint presentation, an idea, or a promise. Rawal understands that he will have to prove the superiority of his approach if it is going to get traction.
The fact that Rawal, his co-founder and COO Monish Suvarna, and founding VP of engineering Pradeep Madhavarapu have plenty of scar tissue from previous adventures in big data (including stints at Yahoo, Cloudera, DataStax, TIBCO, Microsoft and others) builds their credibility.
To that end, Isima has had a handful of deployments at large companies, including a healthcare company, a telco, and a bank, Rawal says. The company essentially is working the edges and trying to prove itself in smaller use cases as a way to garner enough trust for large companies to take a chance with bi(OS) on bigger applications.
“You have to accept the fact that, even in enterprise, it’s pretty messy,” Rawal says. “I call it the Rust Belt of enterprise architecture, and you have to figure out a way to fit yourself into that world with your solutions, and still show the value prop with one analyst working for four weeks.”
For example, one hospital is looking at using Isima to consolidate and present healthcare and patient information. “For doctors, Tableau is too complicated,” Rawal says. “A doctor really wants an iPhone app to say, this patient should be looked at first.”
But Isima isn’t just a dashboard. While it’s not looking to recreate Tableau’s rich functionality, it does provide a place to execute joins and to perform aggregations, which are tasks usually handled by jobs running in a data warehouse or a staging server. By housing the relevant data together, Isima can simplify the process of connecting data sources and handling tasks usually done by ETL products, such as validating, cleansing, and enriching data, and performing joins and aggregations.
“The challenge we’re really taking on is ETL and the storage problem between a transactional system and the warehouse,” Rawal says. “Traditionally you would write an ETL job in Informatica or something, and our view is that we make it less than five clicks, so a product person should be able to do that.”
Ken Denman, who’s the lead venture capitalist with Sway Ventures, which is investing in Isima, has studied this problem for years and thinks Isima has the right technologies working on the right solution at the right time. By consolidating the data sprawl and streamlining the chain of data command, product owners and other decision-makers within companies will be able get answers from data much quicker than they can with today’s data apparatus.
“We have product people who are beating down on these [ETL] guys, going ‘Hey, I need this! I need that!’ And they can’t respond, because it’s just too damned complicated,” Denman says. “So you’ve got to make these guys heroes, effectively. That’s what it’s all about.”
While Isima ostensibly is about simplifying data management and streamlining data engineering, the end result of actually achieving that is opening up new machine learning and AI uses, which today are hampered because the data supply chain is brittle, according to Denman.
“At the end of day, we’re still in the trough of disillusionment around machine learning, deep learning, because it’s so damn expensive and time consuming and the quality of the people you need to get the data organized and right and curated and clean [is so high],” he says. “We’re automating that piece. ….Give us one Python-trained engineer and two-to-three weeks, and we can get you into production. Full stop.”