Virtualization Technology News and Information
Article
RSS
MapR 2019 Predictions: Organizations That Embrace Hybrid Cloud the Best Will Benefit from AI Initiatives the Most

Industry executives and experts share their predictions for 2019.  Read them in this 11th annual VMblog.com series exclusive.

Contributed by Mitesh Shah, Director of Product Marketing with MapR

Organizations That Embrace Hybrid Cloud the Best Will Benefit from AI Initiatives the Most

Organizations are increasingly incorporating artificial intelligence and machine learning into their business processes. Many of these companies are beginning to realize a revenue and profit advantage over their less data-science-savvy peers. Success of these initiatives, however, is far from certain and can be grouped along the familiar lines of people, process, and technology. While all three of these factors are important, the relative value an organization gains from AI and ML initiatives going forward depends on the extent to which that organization is able to develop (and act on) predictions from all their data in an increasingly hybrid world.

People. Management support and sponsorship have proven to be very useful as organizations begin tackling AI and ML initiatives. AI and ML is new terrain for many, and having top-level backing to go explore provides much-needed air cover as data experimenters try and inevitably fail - at least initially. An organization's budding data scientists need to know their efforts are not mandated to yield greater revenues or higher profit margins right out of the gate. Rather, their work needs to be seen as building institutional knowledge in a crucial area that the organization is committed to supporting for the long haul.

Process. Broadly, the process of AI and ML can be broken down into: data exploration, model training, and model deployment. Data exploration is a critical phase here as data scientists will not always know at the start what golden copies of data should be used for model training - if these golden copies exist at all. Instead, data is often spread throughout an organization, and it is up to the data scientist to work with data stewards to understand which data can be used and how. Enterprise data catalogs like Waterline Data Catalog make this process easier. And toolkits like Kubeflow simplify the process of model training and deployment on Kubernetes environments.

Technology. AI and ML models with the greatest accuracy and predictive power are trained by high-quality data - and lots of it. For these models to work best in the long-term, however, they need to be continuously refined based on newer, incoming training data. The difficulty is most of an organization's data - both old and new - is spread across many different locations. It is critical for organizations tap into all of this data, wherever it might be. Creating and refining models based on data generated in the cloud alone leaves out all the data at the edge where, by some estimates, more than half the world's data will be produced in the years to come. Alternatively, waiting for data at the edge to be transferred to and processed in the cloud means losing valuable time in a cutthroat competitive environment. For applications like autonomous driving, the stakes are even higher; life or death decisions need to be made in milliseconds so roundtrips to the cloud and back are non-starters.

When it comes to driving the greatest results from AI and ML initiatives, organizations that figure out how to tap into the data in their hybrid environment the quickest will reap the greatest rewards. This is not to say the people and process described above are unimportant; in fact, they can be viewed as the price of entry. Where organizations will separate themselves, however, is in the degree to which they are able to tap into all their data across edge, on-premises, and cloud environments. Using more of this data means more potential use cases that can be driven by AI and ML. And using more of this data means AI and ML models that can be made more accurate, both upon initial deployment and over time.

Without the right data platform in place, however, organizations will be hard-pressed to leverage all the data in their hybrid cloud environment. Traditional big data platforms built on HDFS, for example, lack consistent point-in-time snapshots, mirroring, and real-time replication, all of which are required to orchestrate data and move it seamlessly to where it needs to be. Traditional big data platforms also have no global namespace or single view of data across disparate deployments.

The MapR Data Platform has no such shortcomings. With support for global namespace, consistent point-in-time snapshots, mirroring, and real-time replication, the MapR Data Platform is an ideal choice for organizations wanting to embrace hybrid cloud to maximize their AI and ML initiatives. With MapR, data scientists at all levels of experience can easily explore and leverage data in their models, wherever that data might be.

The MapR Data Platform is an all-software data platform for AI and analytics that runs on commodity hardware across on-premises, cloud, and edge deployments. Open APIs and support for containerization ensure broad distributed application access and seamless portability of applications across disparate environments. Support for POSIX allows cutting-edge AI and ML tools like new Python ML libraries to run on the same MapR cluster as your analytics. This means no AI/ML silos. Users have access to all data in place from any compute profile thanks to open APIs and container volume plugins.

##

About the Author

 

Mitesh Shah is Director, Product Marketing with MapR. Prior to MapR, Mitesh held positions in enterprise security at organizations including the Federal Reserve and Salesforce.com. Mitesh has a degree in computer science from Cornell and an MBA from The Wharton School of the University of Pennsylvania. You can contact the author at miteshshah@mapr.com
Published Monday, December 10, 2018 7:22 AM by David Marshall
Comments
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
top25
Calendar
<December 2018>
SuMoTuWeThFrSa
2526272829301
2345678
9101112131415
16171819202122
23242526272829
303112345