Taneja Group Analysis
It’s almost 2017 – where is your organization with regards to getting any value out of Big Data projects? Are you still dabbling with endless POC’s, or maybe haven’t even gotten a big data project approved yet?
I just got back (red-eye this a.m.) from MapR’s first ever analyst conference. More on MapR in a moment, but first let me tell you about one of the interesting big data market adoption perspectives I heard there. One of their partners claimed that a big roadblock to wider big data adoption was senior IT decision-makers holding out investment or project approval until they could be assured that any resulting application would be able to be put into actual enterprise production, regardless of business value. In other words, big data adoption may be gated (and throttled to date) by operational support concerns, and not simply judged on the potentially huge and revolutionary business opportunity it might present. It seems too many senior IT folks have been burned by eager early adopters using bleeding edge technologies to develop something super-cool, with obvious business value, but unable to be supported in production due to security, scalability, performance, availability, data protection/DR, actual cost or other enterprise operational concerns.
Actually this is probably the definition of a senior IT person – have you learned this lesson yet? Because that’s an old story – I’ve been guilty of that myself by hacking up crazy shell scripts or coding one-off ruby code just enough to assemble a really useful data result, but far from being a supportable application in production. (Ever have someone proudly show you their game-changing new report based on an impossible to support maze of crazed spreadsheet macros?)
Obviously this is one area where MapR wants to claim the high ground, and show how their uniquely “converged” big data stack can easily provide for those enterprise needs. Thier story starts with the enterprise quality capabilities of its core big data storage IP and extending out as a target big data platform that simultaneously can host and serve many (most?) types of big data analytics, databases, and “computational engines” (e.g. Hadoop, Spark, Vertica…).
MapR, looking for solid positioning and differentation as they soiidify their pre-IPO posture, also aims to be a leading next gen datacenter platform in a wider sense, able to marry both analytics and operational solutions in one stack. It does make sense, reducing the number of moving parts in an otherwise increasingly complex big data stack (if you just assemble open source projects), while also supporting standard developer API’s to help them quickly build what will inevitably become the future implementation base for many (most?) business applications – smart automated just-in-time operations based on big data analytics (e.g. machine learning). In the past I might have called this merging of two worlds “operational intelligence”, but it’s just as valuable to view it the other way around as analytics-driven operations. I’d come up with a new category, but at this point I think its just the future of all apps to grow smarter, more real-time driven.
There is a lot to dive into about MapR’s technical approach, but at a high level if you squint it looks a lot like proprietary software-defined big data storage, pre-converged (but just the software stack) with as many analytical and operational big data solutions as they can. The resulting “platform” (we’ll give them that, but even they recognize that “platform” is an overused and thus increasingly meaningless term) is thus different than the other big data distributions based on an all open source core stack.
Thus, MapR would like you to see them as a software company with potentially large software company margins, not as a services company with a thinner support model margin. They clearly want to differentiate from the Hortonworks example, which is currently not setting a very good precedent for post-IPO value with thir 100% open source “support and services” business model. I’m not an expert in this area, but it seems to me to be not as black and white as all that – for example market leading Cloudera offers some enterprise IP on top of support (and in an adjacent space, Red Hat has been doing quite well by all accounts).
I think the hidden story here is that MapR can be deployed as a type of data virtualization layer (all data is big data at some point) both on prem and in the cloud. It holds out definite promise of enabling smooth hybridized big data architectures, which I think is not just a holy grail, but inevitable — some day soon!
If you’ve run into an organizational roadblock with big data, I’d be glad to chat about it (off the record is no problem).