Horizontal Future
FOCUS OF INFORMATION-SHARING INITIATIVE SHIFTS TO TAGGING OF DATA FOR CONTENT AND LABELING IT FOR SECURITY.
As a Department of Defense research and development program, “horizontal fusion” may soon be a thing of the past. But the legacy of its focus on information sharing among military and government agencies will live on, and much of the work that it started remains to be done.
Department officials initiated the concept of horizontal fusion to accelerate the implementation of the Global Information Grid. As a DoD technology strategy, horizontal fusion focuses on the cross-functional posting and utilization of data by developing functionality, such as federated search and content discovery and delivery, and models such as the Defense Discovery Metadata Standard.
“Horizontal fusion provides analysts and warfighters with net-centric applications and content,” explained Marian Cherry, horizontal fusion portfolio manager in the Office of the Secretary of Defense. “The program was stood up as an accelerant to jump-start the implementation of the Defense Department’s netcentric architecture. We’ve met all the objectives the department gave us, so there is no longer a reason for horizontal fusion to continue.
“We’ve proven that a service-oriented architecture is within the technical grasp of the Defense Department,” she continued. “We’ve demonstrated that you can tag data and find it in a reasonable fashion. We’ve proved out the concept of metadata tagging and the development of a robust, secure architecture that can be trusted.”
The service-oriented architecture (SOA) was selected “as the path to follow to implement network-centric operations,” Cherry said. “The SOA framework provides a design for interoperability in an environment that houses services. Data, tools and applications are all known as services and are attached in an interoperable fashion to an environment where people can find them through a registry.”
Users can then access services such as data and applications and combine them to create new modules that help them in their work, regardless of the network or platform they are on or which operating system they are running.
All this has resulted from the horizontal fusion program’s work. “Now is the time for the rest of the department to implement this,” said Cherry.
But that’s the issue, because despite the provision of architectures, standards and functionalities, net-centric operations will not work without usable data. That is why Cherry urges DoD data owners to get busy tagging and labeling their data, because they risk falling behind the net-centricity curve if they do not accomplish these tasks.
CONTENT DELIVERY PILOT
Meanwhile, the functionality and standards developed by horizontal fusion will be transitioned to the Defense Information Systems Agency’s (DISA) Network Centric Enterprise Services (NCES) area as early-capabilities programs. In that vein, DISA will soon be issuing a request for proposals for the content discovery and delivery capability, with contracts expected to be let in during the first half of 2007.
The program is expected to deploy a content delivery pilot at NIPRNet and SIPRNet sites. The acquisition will be structured similar to the recent NCES collaboration services contract, in which the functionality is outsourced to one or more vendors and DoD pays for the services on a usage basis.
“Labeling data is a concept that is critical in a network-centric environment,” concurred John Osterholz, vice president/general manager for C4ISR at BAE Systems, who previously served as director for architecture and integration in OSD.
“Content is the reason why we are building all these networks,” he continued. “After 9/11, there was generalized awakening across DoD and the intelligence community for the need to capitalize on the content to be found on the various networks.”
The volume of data on DoD networks is astronomical, and the tagging of data is vital to the process of finding mission-critical information. “We don’t need to know all of the data, just like don’t need to know the content of all the books in the Library of Congress,” Osterholz explained. “What we need to find that information is the card catalog.”
That is why Cherry advises that the first step down the network-centric road is the tagging of data for content and labeling it for security. “That is the most important thing anyone can do,” she said, “even if they are not prepared to become full participants in network- centric operations right now.”
Metadata tagging generates Osterholz’s “card catalog.” A metadata tag placed on a data object allows it to be indexed for content. Labels attached to objects provide an audit trail, help with classification of data and are essential for maintaining data security. “The importance of tagging has been established,” Osterholz said, “and appending metadata is the most efficient way to understand what content is available and how to bring it into your own space.”
“One core piece of metadata is the date/time stamp,” explained Cherry. “If that is the only tag on the data and I ask for all data form a certain one-week period, I’m going to get an awful lot of data. But if I’m only interested in information from last week concerning IEDs, for example, and I’m only interested in those that exploded in Basra and I’m not interested in those that exploded on Saturday or Sunday, but only on Monday and Tuesday, then I will get much more data relevant to my search if I have all those metatags in place.”
Cherry’s office acts as a facilitator of the tagging process among defense services and agencies, bringing programs of record to the table to explain the importance of data preparation to net-centric operations. Still, she admits that “there is still quite a bit of data in the department that is not yet tagged and there are a number of challenges” associated with getting the job done. One such challenge involves getting communities of interest to develop a shared data vocabulary, so that common terms will mean the same thing to everyone. Another challenge involves deciding what to label and when.
SHARED VOCABULARY
What used to be known by the arcane terms of data taxonomy and ontology is now referred to as a “shared vocabulary,” according to Ken Pratt, chief architect at McDonald Bradley, which was the lead contractor on the horizontal fusion portfolio.
“Different organizations have different ways of describing information,” Pratt explained. “There must be 17 different ways to refer to an M-1 Abrams tank or an IED. By developing a shared vocabulary, communities of interest are able to describe the same thing in the same way.” This increases the relevance and reliability of data searches.
“We have long been proponents of paying attention to the data,” added Ken Bartee, chief executive officer of McDonald Bradley. “The network-centric strategy has little value to the warfighter in the field if the data layer is not usable.”
McDonald Bradley has brought 40 different defense data sources online, according to Bartee, who added, “But there are thousands are out there that are not online yet.”
Developing communities of interest involves bringing together different defense agencies that all deal with the same function or process, noted John Sutton, McDonald Bradley’s senior vice president for the Advanced Programs Group. Communities of interest can be of wide applicability, such as logistics or meteorology, or represent narrower areas as strike warfare, time-sensitive targeting and geospatial capabilities.
The shared vocabulary that results from the work of the communities of interest takes the form of metadata extensions to the core set of metadata set forth in the Defense Discovery Metadata Standard (DDMS). DDMS is a specification, which, along with various eXtensible Markup Language (XML) schema, is being implemented throughout DoD in order to tag electronic resource holdings.
“Tags represent the content of the resource like a card catalog,” Pratt explained. “Instead of searching through an entire document, you search the metadata instead. This makes for a more efficient search and lends itself to a higher fidelity result.”
“Communities of interest need to sit down and make those extensions happen,” added Sutton. “That way when users look for data, they will get to the piece of relevant information they are looking for and get to it very quickly. Horizontal fusion functionalities like the federated search will not work properly without this shared vocabulary.”
Deciding what data to tag and what to do with legacy data are left to the discretion of the various communities of interest. “Each community of interest needs to give thought to the concept of information sharing,” Cherry said. “From the data owner’s perspective, not every piece of information should be exposed. Sometimes you want to expose raw data, sometimes only the finished product. That is a matter for analysis by the data owner. I’d like to see the system evolve so that a more trusted environment is developed and more data will be shared over time rather than less.”
According to Osterholz, most military organizations are going to take a forward-looking approach to tagging data. “I think it is correct to look forward because of the huge and expanding level of content,” he said. Once that process has been initiated, organizations can consider “strategically looking at legacy data caches to determine which are so strategic that they need to be realigned with current DoD standards.”
Cherry cited one military organization that hit on an innovative tactic to tag massive volumes housed in data stores. “They have all data types registered in a yellow pages type of registry,” she said. “Then they put in a program that tags the data as it is retrieved by users and labels it at the same time. The data is reinserted into the data store as tagged and labeled.”
DATA OWNERSHIP
The enterprise scope of net-centric operations sometimes means that a central authority must wrest control of data from individual data owners. The Marine Corps’ efforts at readying data for net-centric applications has involved “taking data away from those who think they own it,” according to Ron Harris, an IT architecture analyst at Marine Corps headquarters. “The Marine Corps owns its enterprise data, and ultimately it’s DoD’s data.”
The Marines’ intent, according to Harris, is “to build a body of information where the architecture is managed as a whole. We have disparate data sets. At the end of the day, the issue boils down to whether we have a body of data worth passing across the wire.” Just as communities of interest must decide for themselves the scope of their data tagging enterprise, so must they individually decide how they are going to go about the task.
“Each program of record has developed capabilities according to its individual missions,” said Pratt. “Each has its own technology, its own data, and its own means of storing data, so each must develop an appropriate way to map that data. There is no singular process to accomplish this task. Sometimes programs connect metadata to electronic resources and sometimes it is a manual process. Sometimes there is technology available that is able to extract data from repositories and add the DDMS metatags to them.”
Whatever the process that individual communities and programs choose, the experience at the Marine Corps suggests that they implement processes of broad applicability. Individual units that implemented point solutions for data extraction and tagging saw their projects “die on the vine,” according to Harris, because “they were not repeatable processes.”
All which buttresses Cherry’s argument that agencies and programs get organized now with their data projects. “When the infrastructure is more robust, you’ll be prepared,” she said. “The last thing DoD needs to hear is that you need another four or five years to get up to speed.” ♦
