Do It Cheap, Do It Now

But don't worry: There's plenty of hard-core technical expertise onboard in the form of Appian Corp., the collaboration-software vendor and integrator that handles NKO. We were impressed with the project in general, and with the tight coupling of business objectives and system design in particular (see "Sea Change," for background on the project).

However, things weren't 100 percent shipshape: Several days before our site visit, the system experienced a database server outage lasting several hours. When we asked Appian how this could happen, its answer was simple: The provider recommended a fault-tolerant database design to the customer, but the customer decided not to invest in a fault-tolerant database server.

As the big picture unfolded, this became somewhat understandable: There's a definite order of operations and priorities at work here, and NKO is still in its early phases. From a project-management perspective, these priorities are:

• Define the business case: What is NKO's purpose?

• Do it cheap and do it now: This project cannot be expensive, but the team also cannot take a long time to design and deploy the system.• Measure and track: Have quantifiable measures of success, but not necessarily IT measurements; pick the right measures for the maturity level of the system.

• Build to spec: Build an IT system that allows NKO to focus on the first three objectives without regard to specific technology.

Note that "keep it 100 percent fault tolerant from the beginning" is not among those initial objectives. That's a little hard to swallow, but believe it or not, if the business case is right, a project will not languish and die because of a little downtime. NKO proves this.

The Business Case

NKO is the KM piece of the Sea Warrior initiative, which is a component of Task Force Excel, endorsed by none other than the highest-ranking Naval officer, Adm. Vern Clark, chief of naval operations. A key element of Task Force Excel is to improve the Navy by improving its sailors, specifically by encouraging and fostering pervasive and lifelong learning (see https://wwwa.nko.navy.mil/portal/index.jhtml).Retired Capt. Fred Bertsch, functional integration manager for the Naval Personnel Development Command (NPDC), is onboard with Clark's transformation strategy, believing it to be robust and appropriate. After all, it's the sailors that make the Navy able to strike at enemies efficiently and effectively. Bertsch says Clark saw a number of paths to transforming the human force.

"He said, 'You don't have to do it by training and education. You can do it by distribution. You can do it by compensation. You can do it by advancement,'" according to Bertch. "He chose to concentrate on training, and I think rightfully so, because a more educated and capable, skilled force would allow us to more rapidly transform the rest of the Navy."

After this decision was made three years ago, naval training was formally reviewed. A team spent about eight months assessing how the process--unchanged since 1907--could be made better. The team looked at best practices in industry as well as the academic world.

The result was a killer app the Navy calls the 5 Vector Model, a career dashboard that tracks a sailor's progress in five important areas of development.

While Lt. Eric Morris, NKO's program manager, is very clear on NKO's technology benefits to the sailor both now and in the future (when video-on-demand training is scheduled to come online), he is also very clear that NKO is an IT-implemented knowledge-management solution, not an IT solution that happens to feature knowledge management. Never let it be said that Network Computing undervalues testing and technology evaluation, but one pitfall that snares many organizations is paralysis by analysis: By the time they get moving, technology has passed them by and they still haven't implemented anything. Unfortunately, according to Bertsch, the Navy "had a tendency to want to know 100 percent of everything before we take a step forward."In fact, at the beginning of Task Force Excel, the group, which knew that a knowledge-management system was integral to the project, felt frustrated by the pervading sentiment that there would need to be studies and large investments of capital, and that such a system would take many years to build and deploy.

After Bertsch and Morris presented their vision of NKO, the admiral gave his blessing, but no new funding. Thus two imperatives shaped the architecture of the project: It had to be inexpensive, because monies would be taken from other training programs; and it had to work well enough to not only accomplish the new critical business goals, but also prove that taking money from other training activities would not impact the overall effectiveness of the Navy training organization.

Appian started NKO in September 2002, and with Morris as project manager, built the portal in about four months, spending, in total, $4.5 million for hardware; software (components of Appian's Enterprise, its "kitchen sink" product that includes portal management, content management, document management and other software); associated customization; and support and hosting fees.

It's no wonder that the NKO team has a sense of accomplishment. Says Bertsch, "I still have e-mails from people telling me that we were raising the expectations of sailors--it will take 10 years and $15 billion."

Measuring SuccessThe thought right after "Ready, aim, fire" should be, "Have we hit the target?"

The Navy had the right idea about what would constitute the ultimate measure of Task Force Excel and NKO's success: Has the Navy's fighting force become more efficient and more effective?

Unfortunately, like many large initiatives, no significant organizational change is expected in the very near term, so a year out of the gate, it's premature to look for results like "fewer training-mission errors made." However, these numbers are already tracked, and it should be easy to see trends as NKO becomes a part of Navy life. Naval Personnel Development Command is also working with the American Productivity and Quality Center to track this type of data.

"It will be really interesting to see this mature," says Rear Admiral Kevin Moran, Commander of NPDC and director of Task Force Excel. Moran expects that the

5 Vector Model and NKO will significantly improve sailor retention."You won't recognize the U.S. Navy," he says. "Right now our Zone A first tour retention is in the high 60 percent--this will make it even better."

For the moment, the goal of NKO is to get "market penetration," because it can't affect sailors unless sailors use it. Therefore, for the very short term, success is measured by voluntary adoption. By this metric, NKO has been successful--gaining 1,000 users per day.

Other statistics that will be gathered over the next few months include satisfaction with training, as well as the number of "communities of practice" (communities based on specific skills or trades) that sustain themselves--that is, that continue to have active traffic.

Building to Spec

Translating Task Force Excel's business objectives into NKO required that the Navy use in-house portal-building expertise--and take on permanent staff to do a quality job of maintenance and support--or partner with a vendor that could translate the Navy's vision into hard-core tech without losing sight of the all-important business objectives. Appian was the Navy's choice.Myles Weber, an Appian program manager who works with the NKO team, says, "We're big fans of lots of equipment that is smaller. It helps us with scaling up." Instead of buying one huge server, the project team has concentrated on buying several smaller ones and load-balancing among them, clearly playing to the "cheap/now" part of the NKO business model.

Although neither Appian nor the Navy would provide us with all of the technical meat we requested, citing security concerns, they were sporting enough to show us at least some of the inner workings of NKO.

In general, NKO is built as you'd expect a portal system to be: a farm of application servers, Web servers, and--coming soon--multiple database servers. Two Sun Microsystems 880 quad-processor servers do the heavy lifting. The database server runs Oracle8i. A search server currently runs Verity's search engine, but will soon be converted to Convera's search engine for the sake of standardization. The balance of the servers, all of which run Solaris 2.8, are Sun 280Rs: two single-processor Web servers running iPlanet 4.1 and four application servers with dual processors. There is a Cisco Systems load balancer in front of the Web servers, and documents are stored on an EMC Clariion 4500 network-attached storage system (see graphic "Navy's NKO Architecture").

Appian, a J2EE shop, is in the process of converting from ATG's application server to BEA Sytems' WebLogic. The SIPRNet (Secure Internet Protocol Network) version is already on BEA WebLogic; the NIPRNet (Non-secure Internet Protocol Network) version is due to be converted in September. Challenges the NKO team faced included:

• Validation and security: Although administrators can assign access to Navy and guest personnel (for example, we requested guest accounts on NKO prior to our site visit, listing Morris as our sponsor, and our accounts were quickly validated), the main methodology for assigning accounts is a tie-in to DEERS (Defense Eligibility Enrollment Retirement System), an existing database that holds information for all Navy personnel.Weber told us that the system uses Web services to call over to DEERS via XML. "We don't even know what the database is," he says. That's an interesting live example of "the promise of Web services."

At the moment, the authentication mechanism consists of user names and passwords entered via an SSL Web interface, but a Department of Defense mandate requires all externally facing systems to use smart cards by Oct. 1. Weber and Morris are still working those details out.

Smartcard or no smartcard, classified information doesn't ever make it onto the NIPRNet, on which NKO runs. Instead, it uses the SIPRNet, which is a private and physically separate network. This "air gap" makes it safe for engineers to chat about the finer points of nuclear power on SIPRNet NKO. Because the workstation is on a network that has nothing to do with any other network, including the Internet, there's no chance of the latest Blaster-type worm subverting the workstation. All SIPRNet terminals are in physically secure areas, and there is no generic remote access. SIPRNet users need to go to designated locations to access the SIPRNet version of NKO.

Of course, that doesn't mean NKO is off the hook in terms of NIPRNet security: NKO operates as a trusted network in the NIPRNet context.

"To live where we live in the DMZ down in Pensacola, Fla." Weber says, "we have to follow all the other security regs that the other systems do." This translates into being prepared for vulnerability assessments and other spot checks by Fleet Information Warfare Command.• Performance and reliability: Because user adoption and retention, or "stickiness," is part of NKO's goal scorecard, keeping the site responsive is a priority. As far as scaling goes, we asked Weber what capacity planning tool or methodology was used when building NKO. He said that Appian had rolled out a similar system for the Army and knew that its code and systems would scale in a linear fashion. There was, however, a hugely different approach between the Army and Navy rollouts (see "Army-Navy Game,").

"We knew what we could afford," Morris says.

But even though there wasn't money to build NKO as a 24/7 system, Appian and the Navy used tools and techniques to keep the system afloat most of the time. In fact, Weber claims that in the past year, there have been only 30 hours of downtime, 20 of them for scheduled maintenance. And 30 hours of downtime equates to a yearly uptime of 99.65 percent.

Keeping downtime to a minimum on a system that isn't flush with funding has required the team to keep a close eye on resources and to use backup tools that eliminate backup windows: Appian tracks resources with BigBrother and uses Veritas' Oracle plug-in to perform transactional backups.

This goal also requires discipline with change management. "We have development, staging and production systems," Weber says. The code development team uses Ant for code builds and Borland Software's StarTeam for configuration and change management.• Bandwidth: Finally, because many of NKO's constituents are at sea much of the time, one obstacle to a good user experience is bandwidth: NKO's current training materials aren't huge, but large training videos are on the horizon, and obtaining a slice of precious bandwidth at sea presents a challenge.

"Intelligence products, imagery, phones all pull from the bandwidth pie," Morris says. "Nature and the Navy abhor a vacuum, and if I give a ship that had 64K a T1, it will take no time at all for resources to suck that dry."

Accordingly, Appian built caching into the NKO application (using Appian's algorithms). The Navy is going with local presence with preloaded content, then refreshing content when bandwidth is available. This is happening on the USS Mason now, and the practice will expand as NKO grows.

Jonathan Feldman is director of professional services for Entre Solutions, an infrastructure consulting company in Savannah, Ga. He has worked with and managed technology in industries from health care and financial services to government and law enforcement. Write to him at [email protected].

Post a comment or question on this story.Navy program manager, Appian corp.At Work: Ensures NKO's technical capability meets the Navy's business requirements, and manages Appian's consulting team and subcontractors

responsible for NKO.

At Home: Single. Hobbies include rooting for Chicago Cubs. (Believes this is their year.)

Alma Mater: University of Illinois at Chicago, B.S. in information and decision sciences.

How He Got Here:2000 to present: Consultant and project manager, Appian Corp.

1997 to 2000: Independent Web developer and business intelligence consultant

1996 to 1997: Reported market information for brokerage firm at the Chicago Board of Trade

Mouthing Off:

Greatest challenge for NKO: "Addressing every requested change to the system. It's a good problem to have because it shows the Navy is using NKO and heavily relies on it to function on a daily basis."Toughest moment during NKO project: "Picking up the acronyms for Navy organizations and systems."

If only I had a bigger budget for NKO: "Redundancy and failover can always be increased."

Worst high-tech invention ever: "Electric can opener. Using elbow grease is much more efficient."