Cloudera Machine Learning MLOps suite generally available as it aims to manage models, analytics

Cloudera Data Platform has seen solid traction in recent quarters from Cloudera and the hope is that CML MLOps will ride shotgun.
Written by Larry Dignan, Contributor

Cloudera is betting that it can fuel future growth by becoming critical to deploying, managing and governing machine learning models across enterprises and industries.

The company said its Cloudera Machine Learning MLOps suite is now generally available. The effort goes along with its Cloudera Data Platform (CDP) and plays into the company's plan to become more than a Hadoop distribution player. Cloudera merged with Hortonworks last year and set a strategy to manage analytics workloads. The general theory is that Cloudera can be a single pane of glass for multiple data analytics workloads using various Hadoop open source tools.

What Cloudera is doing is launching MLOps to give enterprises a streamlined way to manage machine learning throughout the lifecycle of data. MLOps puts models into production and Cloudera SDX is used for models. Cloudera's plan is to give data scientists, machine learning engineers and operators one unified production line for models. 

Andrew Brust: Cloudera's MLOps platform brings governance and management to data science pipelines

Santiago Giraldo, senior product marketing, said Cloudera built MLOps with participation from a handful of customers as well as the open source community. "We built something that is 100% open source production machine learning at scale," said Giraldo. "What was available for machine learning ops was focused on just the model building and production. You can't skip lineage. "

MLOps lands at a good time for Cloudera. CDP has seen solid traction in recent quarters from Cloudera and the hope is that CML MLOps will ride shotgun. If Cloudera can give enterprises a machine learning platform to standardize operations it can become a cog in the move to artificial intelligence. The majority of machine learning projects don't make it past pilots and experiments.

Features of Cloudera's new launches include:

  • MLOps: Detect model performance and drift over time with native storage and access to custom and arbitrary model metrics. 
  • MLOps: Tracking of individual predictions and ensuring models are compliant and optimized. 
  • Cloudera SDX for models extends governance to machine learning to track, manage and understand models across the enterprise with cataloging, lifecycle lineage and metadata in Apache Atlas. 
  • Cloudera SDX for models has security for REST endpoints. 
  • Integration with Cloudera Data Platform and built on open source standards.
  • CML MLOps is available on Cloud Data Platform for both Microsoft Azure and Amazon Web Services.

Primers: What is AI? | What is machine learning? | What is deep learning? | What is artificial general intelligence?  | Free PDF download: Managing AI and ML in the enterprise 2020   

Alex Breshears, senior product manager of production ML at Cloudera, said the opportunity for the company and MLOps is to deliver a platform that can be the "first step in solving nebulous question of where this data model came from." 

Integration, leadership changes

The launch of CML MLOps comes as Cloudera appears to be hitting its stride a bit. The company's most recent fiscal first quarter was better than expected and customers seem to be embracing Cloudera's CDP and strategy overall.

Wedbush analyst Daniel Ives said following Cloudera's first quarter report that the company is showing positive momentum with its CDP platform, which launched in September. Ives also noted that the integration of Hortonworks and a revolving door of executives seems to be behind the company.

Indeed, Cloudera since acquiring Hortonworks replaced its CEO before naming Rob Bearden chief in January. Acquired Acadia Data, launched CDP and delivered three quarters of improving results following stumbles and fierce competition from cloud giants.

Ives noted:

Cloudera took a much-needed positive step in the right direction for the company after a challenging post Hortonworks acquisition period. The company (and its investors) have been through a hurricane-like storm over the past year as Cloudera has tried to successfully integrate the Hortonworks deal, while navigating a choppy Hadoop market. That said, with a product and sales integration that now appears be in the rear-view mirror with improving customer demand and a number of new products.

Ives argued that Cloudera can now be a player in hybrid cloud deployments and improve its execution.

Bearden, co-founder of Hortonworks, on Cloudera's first quarter conference call said CDP is showing momentum and AIops can be a key market. He said the integration of Hortonworks and launch of CDP was phase 1 of a multiyear effort.

In Phase 2, we will transform from a mostly on-premise enterprise data management vendor to a true hybrid multi-cloud data platform company. We now enable modern enterprise data architecture and manage the entire life cycle of data for multifunction, multi-cloud use cases. And Phase 2 will be characterized by 4 fundamental things: first, newer product innovation in the form of additional cloud native services offered as part of CDP Public Cloud; the introduction of CDP Private Cloud; an increased emphasis on the Edge and real-time streaming opportunity with Cloudera Dataflow as the service. We intend to amplify the comprehensive nature of our solution set and deliver the entire portfolio of assets that we brought together with the merger and have since deployed. Our CDP Public Cloud services will reflect our competitive advantage in addressing the full life cycle of data, being able to manage data from the point of origination, process it in real time, report on it, direct the next action using machine learning and leave the workload to the optimum environment for use case performance and cost.

Cloudera's bet is that it can manage the data lifecycle across the enterprise and Cloudera Machine Learning MLops is a step in that direction.


Editorial standards