Trip Report

DARPA/ISO Integration and Transition Off-Site Conference

Craig Thompson
Object Services and Consulting, Inc.
 
Arlington Hilton
21-22 October 1997
 

Executive Summary

The purpose of this two-day meeting was to review DARPA ISO application programs (JTF, JFACC, ALP, DMIF, BADD, Genoa) looking at their architectures and how they begin to comply with the emerging AITS architecture framework. Around 150 DARPA ISO architects and project members participated in the meeting.  All briefings are available on the web at http://www.les.mil/aitsjpo/darpa_iso_ipt/oct97/index.htm.  

My role at the meeting was as one of a small review team reporting to Dave Signori (Deputy Director of DARPA ISO), who is leading this architecture integration and technology transfer effort. This is my trip report, which just consists of meeting notes from Day One (presentations) and Day Two (breakout sessions and meeting summary), probably not very coherent to others.

Separate documents contain

Day One

Introduction, Dave Signori

This was the second Integration and Transition Off-Site Conference. The first was in February 1997. A third Integration and Transition Off-Site conference is planned in April 1998 with an Information Superiority Demonstration in August 1998.

At this meeting, there will be three working groups:

The central concept of JCS Vision 2010 is information superiority. Advanced Battlespace Information System (ABIS) is a guiding study on how to translate this into an evolving system of systems view. The ISO game plan, that DARPA Director Larry Lynn agreed to a year ago, is to develop a pilot system, with global information management, integrating a set of applications including the programs already established and link them into a common framework, guided by a common architecture. The seed architecture was JTF's component architecture, generalized. Also, to establish a virtual collaboratory and integrated demonstration. The transition path is to expand the Joint Program Office (JPO) which transitions technology from DARPA to DISA.

FY97 highlights: Improved program definition for BADD, DMIF, DDB (new), JFACC (July demo), DMIF/BADD (Aug/Sep demo), ALP/Genoa (Sep/Oct demo). Strategy for integration and transition of components in FY98/99 - form partnerships so JTF servers are now spread out into other programs. New commitments to engineering. Working on object-web (they don't mean the web) and object model. In working groups, we will try to refine the AITS architecture. Many kinds of products are emerging.

Introduction, Marv Langston, new Director DARPA ISO

Marv Langston is the new director of DARPA ISO. He was chief information officer for Navy staff.

The goal is to harness information; we are at the beginning of the information revolution. Marx's material conception said technology of the times drives the rest of the times. DARPA budget is 70% of DoD Science and Technology $ so we must be responsible. Charge is to develop a new order of things. Metcalf said the power of a network is equal to the square of the power of the information and people connected together. So our applications cannot be isolated. Langston is behind the notion of an object architecture.

AITS Architecture, Signori, Deputy Director DARPA ISO

Nice slides.

Much JCS activity is to flesh out Vision 2010 that DARPA fits into. Many ACTDs, mainly aimed at battlefield and not logistics transitions. DMSO has modeling and simulation and DARPA has a lot of the rest of funding. Went over ABIS capability Framework (at best a 30,000 feet notional architecture). Grid level supports end to end connectivity and security plus data access.

New ISO initiatives include Command Post of the Future, Dynamic Database, Integrated Battleforce Management.

Planning more participation in OMG - good. Model for DARPA is components coming out of programs transitioning from DARPA programs to APII to DII LES to DII.

Operational architecture - description of teaks and activities, flows to support military. Technical architecture is minimal sets of rules governing arrangement, interaction, and interdependence of parts or elements. At work identifying research holes.

What's different? Virtual teams, better integrated across function, level of command, time, more agile organizations, better coordination, explicit management of uncertainty, much more. Signori is a master craftsman of foils and presentations. Capability levels from brittle to agile with people operating at many levels.

Want component reuse. Lots of efforts in parallel. Expedite ITO to ISO transition. Reviewed 89 projects sorted into readiness and investment area. Mentions Habanero, Berkeley Digital Library, Caubweb, MediaNet, Integrated Digital Library, built into Mosaic … - being transitioned into JPO environment. This was a pilot.

Looked at relevance of JTF servers to programs (matrix). Looked at interface and implementation maturity. Similar picture in data services showing ontologies, repositories, Infosleuth, …

Lead architects for ALP, …: Steve Milligan, Jay Mork, … slides fly by.

Approach to evolution. Always dealing with heterogeneous info, view of next few years, (0-2, 3-5, >5 for research) Iterate for stability and time components. Capitalize on COTS, accommodate Java and DCOM+, other DARPA projects. Actively influence DoD architecture JTA/COE. Concerns about scalability of AITS. Agents effort. Talks about product lines and DSSAs.

Cells federated of object web, model, map, plan, situation assessment, agent applications on assured data floor and adaptive comms. With ops, Intel, Logistics links connecting the cells. If only foils were architecture - well, they sort of are.

Architecture key concepts - specialization of services hierarchy, clusters, DSSA, applications framework, OO, just in time logistics. Shared infrastructure with programs built on top.

No work yet on: Missing framework for configurable user environments. Distributed systems management, information logistics.

Implementation guidance: started with DII/COE, added OO, minimum qualifying tier, higher level tiers. Capability tiers. OS = Solaris and NT, languages = C++, Ada, Java for user interfaces only [audience thinks this is now short sited, I agree.]

Challenges and opportunities: HTTP and web browsers, CORBA/ActiveX and Java, CORBA performance, JTF web server, emerging OMG services, security definition, mobile agents, 3D visualization and collaboration. Some of these are not resolved. Real-time need to track updates are 20X faster than current ORBs.

Covered list of things for architecture working group to talk about.

Interoperable Data Presentation, Todd Carrico, tcarrico@darpa.mil

three topics. SPAR, under Tom Garvey. AITS WarPlan kernel of data interoperability. And Common Object Base.

SPAR is Specifications for Plans, Activities, and processes. DARPA and Rome. Core group being led by Austin Tate. Specialisms Experts Panel. Formalization review panel. Not just plan but possible futures. Much past work on generalized plan representation. Aiming for a reference implementation. Several representations which are different views of same model including formal English, OO, ontology, ER and language. Core document and set of representations like KIF, …

AITS WarPlan. Specializes SPAR for Log, Ops, and Intel, which overlap. Common operational picture is the intersection. Focus is initially on the set of objects exchanged not all objects until 2001. Wants to put this picture in the middle and have all other parts connect to it. Need unified scenario to do good job. ARPI (tech base planning and decision aids) program provided first cut of this or SPAR.

AITS Common Object Base/Schema/Ontology. Vision is richer representation. Combination is KBMS, ontology, repository, federated ontology. Java ontology. Creating central repository. Data Services Vision. Might want to invite DISA/SHADE folks to ease transitions.

JTF/ATD and AJP/ACTD, Ref Delgado, rdelgado@darpa.mil

Described operational puzzle with various components. JTF providing infrastructure. Single plan representation across echelons and single plan with multiple versions. Moving to composable object web servers. Corbus web server.

Want to use Naming, Security, a few other services from CORBA. CORBA extra services: persistence, web, policy, versioning, trigger, replication. Now moving to Java-Enhanced Data Services. Corblets. Cache manager. Gunning OKBC (HPKB version of ODBC). Map server moving to JMTK. Supporting IDL and also C2Fine representations.

Many echelons and applications

JFACC, B. Plebanek

Developing decision maker environment, not just for domain of air ops. Need common plan representation - we have intel, ops, campaign, mission, … plans. So JFACC is linking all parts together. OO and web oriented. Moving away from anchor desks to cross functional planning schemes. Cyclical plans - 24 release. Inefficient. Should be in a dynamic environment. Give decision maker max time to evaluate the changes. Instead review multiple options than specific plans. Using case based reasoning from previous situations.

Karl DeFranco (LOGICON) took over and covered the details. Multi agent planning. Dynamic adaptive managed workflow. Synchronous and asynchronous planning. Coalitions. Reducing manpower.

JTF objectives - air and maritime and land objectives - tasks - resources - situation model - tasked activities and projected activities.

JFACC related programs include ICV, HPKB, I*3, BADD, DMIF, AIM, IO ACTD, P&DA.

ALP, Brian Sharkey

Logistics covers in-storage, in-process, in-transit, and in-theater. Joint Logistics Technology Office formed this year to house JL ACTD. Focused on mature technology. ALP focused on revolutionizing logistics technology base. Trying to highly automate the process. High detail, continuous plan generation. Architecture is different from AITS - primarily agent based. ALP feeds into Joint Logistics ACTD, which feeds into DDJPO, then into GCCS, a heavily web-based traditional target (no, its not says Janet Putman, sitting next to me). Tons of org charts. Technical architecture is evolving in JL ACTD. GCCS provides world wide access to data sources. DISN backbone government Intranet based on SIPRnet (I thought this was dead). Reliable communication in remote regions via phones, pagers, … - use phone to NL query remote DBMS.

Tod Carrico then talked about convergence of ALP Vision and … Want one architecture where different organizations can have different structures, assembled into larger organization dynamically. Cluster management. Allocator plug-in, Task Expander plug-in, Assessor plug-in, ALP Data Management plug in. Clusters combined to become societies.

Distributed heterogeneous data sources. Very large number of distributed objects. Society of clusters. Agent based architecture. Targeting level 6 and maybe 7 in DII/COE.

Using Visage for common visualization.

DARPA/DISA Information Superiority Demo, Don Eddington

Integrated Force Management in ISD-98. Functional scope: Genoa, DMIF, Situation Services, BADD; JTF ATD, Integrated Force Management JFACC-after next, Focused Logistics via Joint Logistics and Advanced logistics Technology, Information Assurance IA Services, intrusion detection and coalition release. Pilot services are early release of not quite hardened technology. Notional sets of ISD pilot services: Once these services are on the network, they are there 24 hours a day. Starting with the first sets of services first - aiming for integrated demo by August 1997. Data services are available to be used. GCCS mainly implements deployment. Not sustainment, several other areas. People are sharing data, visualization, and collaboration. Shows picture with virtual situation room (collection of virtual rooms).

Lessons learned: you find out a lot by integrating things, crystallizes things.

Course of Action Analysis, Dell Lunceford, dlunceford@darpa.mil

Looking at first cut at commanders intent for commander in field. Hasty vs deliberate plans. FM 100 and FM 101-5 (chapter 6) are Army documents on missions planning.

BADD, Bob Beaton

Profile Editor driven off web that downloads the information management tools. Policy editor with tailored extensions. Access and Resource policy management as in comm. bandwidth. Tailored extensions to common infrastructure. Have detailed APIs. Register products via advertisement editors and gatherers. Catalog manager and broker. Delivery management. Site management, QoS policies generated, site manager, transfer planner, communities of interest. Whole delivery management is out of band. Apps are not information management aware. Http://38.231.7.2/badd pwd POC: Established a number of liaisons and determined what's involved for more. Products by FY98 packaged to DII COE level 5 include Policy Editing Tools, resource management service, info flow management service, several more. Issue of integration of BADD, IDM and Comm. Server they are behind on.

Genoa, Brian Sharkey

Program is four months old.

Current chaotic process at NCA level to get relevant data associated with crises and creation of courses of actions - as quickly as possible. Need to create corporate memory so lots of information about past situations is not lost regarding lessons learned.

Process: Query (many unclassified and classified data sources) - Receive data - prepare data - visualize -analyze - review - assess - generate options - prepare briefing - decision makers involves collaboration and sharing information. Using UML.

Tightly coupled with a lot of government agencies.

Showed architecture CrisisBrowse, Visage, Virtual Situation Book, Multimodal interface, Genoa desktop for configuration, Trigger server, workflow manager, collaboration server, pedigree server, situation server, web server, logging server, object management, comm. server for bandwidth, structured argumentation, evidential reasoning, data server, news on demand, DCOM and CORBA, map server via JMTK TBD,

Crisis templates. Crisis information packages. Ontology definitions. Current schema is narrow, needs to be broadened. Will acquire Ontolingua, Ontotharus, IDL Generator. Will use segments that are appropriate.

Will user data server, trigger server, comm. server, JMTK or some map server. Provided by Genoa meeting transcription and Summarization, Collaboration tools, multimodal tookit. PlaceWare is first candidate for collaboration environment, Java based, server, toolkit, environment, spin off of CVW, commercial, backing of Sun. They are new and small so Genoa can influence. Auditorium is a collaborative environment where one speaks to many. Genoa is web-based - whole demo will be available over the web so use DARPA homepage as demo environment.

Issues: data server cannot access web based data. Server documentation is vague and IDL interfaces are not available.

DMIF, Steve Flank

Program is five months old. fusion engines, product finishers for mapping to end users, encapsulation for legacy fusion engines (like wrapping but include performance info). Tenney is associated with DMIF. Lots built on DII COE level 5 segments from Army ASOS program. CORBA is implementation glue - using CORBA 2.2. C++ preferred, some in Lisp, user interface in Java. ORDBMS currently Informix Universal Server. All modules delivered as DII COE segments. Hiding fusion engine implementation language. Using object-relational, not just object store. Using ISI SIMS. Code built segmented from the beginning. Incorporating TIPSTER NL, HPKB, STOW. BADD and DDB integration planned. Tactical Terrain Analyzer - reached agreement to use this. DMIF is architecture enthusiastic. Future products: fusion management, inference model server, pedigree server (software tag comprising time, source, quality, process history). Tech transition cost 8 months and countless messages to receive feedback on how to be compliant. So its hard to get feedback from JPO. Cannot use Dataserver, doesn't use OQL, doesn't tap into Intel side or do schema translation. Similar with access to model server. Concept to query mapping. Issue is real-time performance issues for a Real-time CORBA.

Signori Closing Remarks for Day One

People have made progress on architecture since the February Off-Site meeting. Described logistics for Day Two breakouts.

Day Two

Architecture Working Group

About 50 people participated in this morning's session.

Mid-Term Architecture, Rick Hayes-Roth, Teknowledge

Holes - see Rick's foils A critical thing Rick is trying to do is mapping foils to a consistent architecture where foils normally are not mapped by more than words.

Architecture made from components reduces useless diversity that does not add value. Rick has a handout and AITS Reference Architecture paper that needs to be reviewed.

Ideal architecture is complete, consistent, correct, and current (4Cs) - but we are almost always out of 4C due to resource limitations so we need to manage. Global information base and world model composed of enclaves of function composed of people and software including agents arranged in cross functional hierarchies. DSSAs for appropriate product lines

Near-Term Architecture, John McKim

This is the 0-3 year architecture.

Areas of importance: data services, management services, …

ORBs Orbix, Visigenics, TAO, JavaSoft's RMI IIOP free ORB, Orbix ActiveX Bridge - multithreaded ORB and IIOP.

Too much emphasis on Publish and Subscribe as a mediator mechanism.

Lots of busy slides as mapping of rubber hits road. Trick is now whole systems show up in little boxes.

Implementation Guidance

They ask for updates.  Recommend wider use of Java than UI.

Long-Term Framework

No notes on this.  Was this related to Rich Ivanetich's High Level Framework?

Functional/Operational Architecture, Stein

Generic Functiona/Operational Architecture. Matching Generic C2 functions and new operational concepts like ABIS, JV 2010, program vision statements. What's different, measures of merit, create overarching strategy.

Tactical Communications Architecture, D. White

Most end users are wireless and have strong constraints on them, lightweight. Want more mobility, have movable capability but takes long setup. Why not a commercial solution? No protection against jamming. UAVs/global hawks can provide warfighter's Internet elements. Single beam covers 300km2. CAN = air comm. nodes in planes. Expect 10 millisecond delays on the satellite or plane-based systems. Many intermittent users. Recommends minimizing putting a lot in the field. The problem is uplink limited.  Look at exceptions.

Issues/Priorities/Directions

Signori outbrief on Architecture Working Group

We reviewed mid-term and near-term architectures. Six holes identified, including one called Killer Apps, a message to the program but not regarding the architecture. Genoa is providing user configurable architecture so start with that. Potential Implementation Guidance: Java-based servers, COTS licensing footprint, use Lisp, address www, host on NT, need ORDBMS, DCOM, require rationale for out-of-compliance approaches.

Commercial communications not jam resistant. Application design and database locations need to consider communications latency and bandwidth.

Review Team

Other changes coming

Shared Object Base (Schema) Working Group

They identified a bunch of areas related to the ISD 98 demo and which programs are working on that topic, then an editor (person) to define the ontology for that area. E.g., constraints, collaboration, security, workflow, blue_order_of_battle, pedigree, terrain, visualization, space, time weather, event, … then decided mechanism to capture scheme - Java, c++, UML, ODF, Rose, IDL, Loom, ODI, RMI, agent, due to language, representation, case tool, ontology, db vendor, transaction mechanism, comments, … ics.les.mil u:C3Ischema p: … apease@teknowledge.com. Telcon 7Nov97. Confusion of terms schema, ontology, model, knowledgebase, … Bill Fabian is leading this. Others: Glen Tarbox, Rob Peace, Jon Weiss, Adam Pease, Todd Carrico.

Integrated S Demo ISD 98 - Demo and Transition Working Group

Taking a functional thread approach. Identified data sources (e.g., MIDB, GSORTS, JTAV, JPAV). Services delivery schedule. Did surface analysis of which functions and from which groups.

Issues: adequate funding, come as you are, schema free, multilevel secure, component available ASAP, refine schedule, validate OPS threads against JWID, PMs assign AOs to ISD 98 project.

Final Comments, Marv Langston

Thank you for coming.

Plans to chair the system engineering board himself. Focus groups will have start and stop mechanism. The trick is to balance continuously between building a big system versus making componentwise progress.

Tack down names, dates, and actions associated with progress within two weeks. Demo is good idea, story line or logic flow is good idea. Our business is information or to inform those who need to make decisions. Our job is to make big changes in the way we do business.

Excuses for failures in past are gone due to single funding source, ISO.