Earlier this month, Teradata unveiled a new marketing campaign to reposition the company to go beyond IT in addressing the business. The company, which formerly positioned itself with variants of the theme of the world's biggest, fastest data warehouse is now positioning itself with a stop the insanity-style tagline, "Stop buying analytics. It's time to invest in answers."
Have we got your attention yet?
It's not surprising that rebranding and changing the message was the highlight of the company's annual customer event last week, Teradata Analytics Universe, which itself is a rebranding from the event formerly known as Partners. The company has just moved headquarters from Dayton to San Diego, where it's likely to draw more of the talent it needs to bring the company forward. There's a visionary future CEO in the wings who's putting his stamp on the company. And there are the modest revenue gains, coming atop a transition to subscriptions, that are starting to put the business on a sounder footing.
So, it seemed like the purpose of this year's Universe event was getting its most loyal customers to go back to their bosses and pass along the more strategic messaging, with the goal of reaching the business execs who write the checks.
Nonetheless, under the hood, there are a number of course changes underway designed to push Teradata on a new path that's driven by convergence between on-premises computing and the cloud. We're speaking of the emergence of what was called the Teradata Analytics Platform, that, given the company's new tagline to stop buying analytics, has been rebranded Teradata Vantage. And that was the one product announcement that happened at Universe: Vantage is now generally available, including packaging many of those R and Python functions that we reviewed back in the spring. On the roadmap are design features to incorporate Spark, TensorFlow, and other custom analytics engines.
OK, there's the obligatory AI angle: With R and Python comes support for extending its analytics libraries and functions with machine learning, graph, and deep learning. Teradata wants you to tap the power of R and Python with the performance of in-database computing. Check.
We won't go as far as terming Vantage as part of a backlash to Hadoop. But with R and Python package in-database support, it is vying to grab back some of those workloads that would have been moved, or pushed down via QueryGrid, to Hadoop from Teradata. But Teradata needs to get better control of the narrative here. Depending on who was speaking, Vantage is either picking up from the failure of Hadoop or joining it in a diversified universe... err, ecosystem. Given the similarity of Teradata and the soon-to-be-combined Cloudera/Hortonworks, for targeting the top 500 or 5000 organizations with the toughest analytic problems, the perception of rivalry is understandable. Hopefully cooler heads will prevail in a situation that is definitely not zero sum
Under the surface, there are even more fundamental changes brewing. Whereas the software-defined IntelliFlex architecture was previously positioned as the destiny of Teradata infrastructure, the draw of commodity infrastructure -- spurred on by the cloud -- will likely prove a big disruptor. There is just so far that Teradata can resist tectonic forces.
Is IntelliFlex destiny or a waypoint to commodity infrastructure? We believe that eventually IntelliFlex will be as "dead" as the mainframe -- it will still be needed for extreme scenarios, but we expect that the increasing power of commodity infrastructure and the scale of the cloud will render this equation moot for many customers. And that sets the stage for Teradata to deconstruct itself toward becoming what it has talked about for many years: becoming a software vendor designed for a world where the architecture is converged for on-premises and cloud deployment.
But getting there requires much more than buzzword-compliance. There are many reasons why Teradata or its customers won't cross the Rubicon with the next product rev.
IntelliFlex introduced software-defined infrastructure to overcome the limitations of Teradata's classic appliances. Instead of buying one box designed for interactive query, another designed for high concurrency, or others designed for IOPS- or compute-centric workloads, IntelliFlex was a box that could be configured to the workload via software.
But then came the cloud. Teradata initially had its own data centers for powering its IntelliCloud as-a-service offering, but the momentum of new deployments are going onto AWS and Azure where it has to work with the hardware of its cloud hosts. That means working with the compute instances and interconnects offered by its hosts, as opposed to the custom silicon and high-speed InfiniBand of its own appliances.
We're not trying to say that Teradata has been pushed around by its new cloud partners to put up or shut up with what they have. Instead Teradata has worked with the cloud providers to develop compute instances better able to handle the extreme requirements placed on them by Teradata workloads. And so, while the initial offering of Teradata IntelliCloud for AWS, for example, is centered around m4-class general-purpose compute clusters, more optimized offerings running on the next-generation m5- and m5d-class instances, backed by networks up to 14 GbE, are on the horizon. While that's not quite InfiniBand territory, it's getting there. And next year, we expect that Teradata will match Snowflake, Azure SQL Data Warehouse, and Redshift (Spectrum) in adding direct access to cloud object storage -- Amazon S3 and Azure Blob store.
Teradata is purposely targeting high-end analytics customers and not trying to compete with the likes of Snowflake. The rationale is that Snowflake, Azure SQL Data Warehouse, and Amazon Redshift are not designed for the types of high-concurrency, highly complex workloads (e.g., SQL queries involving up to 8 or 10 table joins) requiring optimization and workload management, and therefore, are not really competitors. But some Teradata customers are either considering offloading their data mart-style workloads to these fast-growing cloud data warehousing services or refactoring those complex SQL queries to a level that those offerings could digest. Could complex workloads actually be broken down into a bunch of much simpler parts? The proof will be in the pudding.
One Teradata customer that we spoke with had a wish list for IntelliCloud to offer more granular consumption options or tagging of workloads. At first glance, that sounds like a defensive response: offer an apples-to-apples comparison to what Snowflake charges. But in actuality, meeting the competition with workload-specific billing could open new opportunities for Teradata IntelliCloud to actually expand its target market of workloads by adding more flexible tiered or spot pricing that could keep IntelliCloud clusters fully utilized with new workloads that otherwise would have been priced out.
Likewise, the march to commodity hardware is a work in progress. The latest version of Teradata supports a software-only version that is installed on x86 hardware using VMware, with an upper limit of 32 physical node clusters, or 128 virtual nodes. While that might sound sizable, it would only accommodate half of the 800 TByte data warehouse that one IntelliFlex customer that we spoke with actually runs. Currently midway through a three-year contract, that customer is hoping that Teradata on commodity hardware will be further scaled by renewal time; we wouldn't be surprised if Teradata meets that goal.
This month, Teradata has changed its branding, messaging, and positioning. The transformations under the hood toward converged cloud and hybrid architectures are works in progress.