While hundreds of AI initiatives and programs are underway within the Department of Defense, many face new and diverse challenges when it comes to operationalization. Selecting a solution and putting it into practice is certainly not the same task, creating challenges that span both organizational and data facades.
For example, the emergence of hybrid and multicloud architectures presents many with data integration, access and management challenges as agencies seek to leverage data assets that span both on-premise and hosted solutions.
This might be why GSA Data Center and Cloud Optimization Initiative Program Management Office recently launched a Multi-Cloud and Hybrid Cloud Guide to help agencies make better decisions about cloud architecture. Further complicating matters is that the DoD is facing a groundbreaking December price of the Joint Warfighting Cloud Capability procurement†
John Sherman, Pentagon Chief Information Officer, describes the JWCC as a “multi-cloud effort that will provide the Department of Defense with enterprise cloud capabilities across all three security classifications: unclassified, classified, and top secret all of the continental United States.” States to the tactical lead.” At the end of this possible five-year tender, the DOD plans to launch a full and open competition for future multi-cloud acquisitions, until then, DoD data scientists may be forced to work in silos because connecting to live data cannot. is always possible.
How technology can help
Many government agencies are overcoming these challenges using technologies such as data virtualization to implement a logical data fabric approach capable of ensuring reliable data access and sharing. Data virtualization is a modern data integration technique that integrates data in real time without having to physically replicate it.
Data virtualization can seamlessly combine views of data from different sources and feed AI/ML engines with data from a common layer for data services. Using data virtualization, AI teams can work more efficiently and collaborate more effectively because the technology provides views of data rather than replicating it.
Not only does this save access and storage costs as it provides a unified data access layer, it also enables stakeholders to implement governance controls from a single point in the department. Creating this “single source of truth” is one of the most valuable features of unifying corporate data using data virtualization.
This combination of enterprise AI, multi-cloud architecture and data virtualization is being used by many government organizations to leverage data more effectively and take advantage of the cost savings of the cloud. Together, these technologies underscore the fact that digital transformation is not just about technology, but about using technology in the most intelligent way and bringing value to data science teams and internal and external data consumers.
As the competition for market share among the major Cloud Service Providers (CSP) promises both better value for government and access to a wide variety of AL/ML tools to achieve better, mission-specific results, there may be conflicting stories about CSP — and which AI/ML technology — is best for a given mission.
In this environment, a logical data fabric is rapidly emerging as the technologically elegant solution to the chaos of multi-cloud computing, as it simultaneously makes the best features of each CSP available to users.
Defined by Gartner As a design concept that serves as an integrated layer (fabric) of data and connecting processes, a data fabric leverages continuous analysis across existing, discoverable, and derived metadata assets to facilitate the design, implementation, and use of integrated and reusable data across all environments, including hybrid and multi-cloud platforms. This pure-play data structure enables the best of all commercial CSP offerings without vendor lock-in and in many cases proves to be the government’s best response to address these challenges.
Addressing the ubiquitous silo data of the DoD, standardizing and improving its quality and access, should be a prerequisite for having the data needed to train algorithms for many defense applications. A logical data fabric approach that includes data virtualization promises to provide a means for rapidly collecting, processing and using information from the various data sources of the DoD. It also ensures that a developed AI/ML model in a silo is still relevant to live data and can accelerate better data flow and data access across the entire AI operationalization cycle.
The more data AI/ML models receive, the more they learn to make better and more accurate predictions that the DoD needs for mission-critical decision-making. However, extracting data from multiple sources and then replicating it to a central repository is an old and inefficient way to access data. The process is still prevalent throughout the federal government and often results in most of the project time being spent on data collection and preparation tasks.
With the DoD touting their growing capabilities in artificial intelligence and machine learning technologies, integrating data to power disparate AI/ML models with their growing data science teams is still a key undertaking. Enterprise AI, leveraging a logical data fabric layer, overcomes these challenges. It can act as a central hub for data science teams between different AI/ML systems, reducing the need for data duplication, enabling highly sophisticated AI/ML initiatives with improved operationalization and accelerated timelines for faster time-to-production.
Bill Sullivan is Vice President and General Manager, US Federal at data integration and management company Denodo.
Have an opinion?
This article is an opinion piece and as such the views expressed are those of the authors. If you would like to comment, or have an editorial of your own that you would like to submit, please Email C4ISRNET Senior Editor-in-Chief Cary O’Reilly†
Want more of these perspectives sent directly to you? Subscribe to receive our Commentary & Opinion newsletter once a week.