An Approach to Replicated Databases for Robust Command and Control

This report presents an approach to making future command and control data more timely and robust through the use of replicated distributed data management techniques. The types of command and control data that are addressed in this report include "situational awareness" data needed by Army tactical commanders from platoon through corps, including mission plan and progress, represented as machine-processable operation orders (OPORDs), enemy situation, and friendly situation and status.

The concepts presented in this report were developed in answer to common objectives in two projects, a Distributed Databases project for the Advanced Research Project Office (ARPA) of the Department of Defense (DoD), and a project addressing the Management of Information for Joint and Combined Operations for the Army Office of the Director of Information Systems for Command, Control, Communications, and Computers (ODISC4).

Motivation

The report describes a conceptual solution to the problem of giving warfighters near-real-time common pictures of the battlefield for situational awareness in spite of communication delays due to jamming and congestion, network partitioning, node failures, and other hostile actions that might interfere with the rapid dissemination of data.

The ARPA sponsor asked us to explore a replicated database approach, assuming future technology including gigabyte networks and memories, terabyte secondary and archival storage systems, and multiprocessor and distributed system architectures, to see if these technology advances could be used advantageously to support C2 robustness. The main issue is the conflict between the need for data in near-real-time everywhere versus the need for data consistency everywhere. Delays in transmission (even normal transmission operations) will mean that replicated data cannot be guaranteed to arrive everywhere at the same time. Given that timely availability of data is extremely critical to C2, the problem becomes one of managing data inconsistencies across C2 units in such a way that distributed data will eventually converge toward consistency at all the C2 units.

The objective of the Army study effort was to support the current Army future-looking endeavors such as Army Force XXI, the Army Digitization Office (ADO), the Army Battle Command System (ABCS), and the Army Enterprise Strategy (AES).

The following quotation is directly related to the issues of this study: ". . . despite advances in information technology, commanders, leaders, and soldiers will never have perfect knowledge of the operational situation surrounding them. Yet, due to the pace and complexity of future battle, commanders, more so than in the past, must accept uncertainty and not hesitate to act instead of waiting for more analysis or information." [TRADOC PAM 525-5, 1994, p. 3–4.]

Approach

The focus of this study was to examine data asynchrony assuming full replication of dynamic command data from battalion to corps. Partially replicated data were assumed for units below battalion and the study does not address how to achieve consistency, timeliness, and robustness when data are only partially replicated. The study does not address communication technology and media, or computer and security technologies, except for noting the context in which topics may become issues. For simplicity, it has treated communications between units as point-to-point communications, though other alternatives are possible and are addressed in Section 1.

We discuss two different ways the Army sees information being managed and provided on the battlefield. One is through nonhierarchical information carousels (organized according to functional areas such as logistics or intelligence) and the other through hierarchical and horizontal distribution of dynamic combat data. Our study is focused on the use of data replication to manage and provide dynamic combat data, though we provide an information carousel example in the management of intelligence data. Our design supporting the replication of dynamic combat data requires that a Knowledge Manager for a C2 combat unit (KMC) exist for each C2 unit (e.g., battalion, brigade, division, and corps). The KMC manages all data unique to that unit through the Army Common Operating Environment (COE) software augmented by any unit/mission-specific software including decision-aiding software. The KMC can be thought of as an automated assistant to the unit commander or staff for carrying out the unit's mission. Changes to the unit's database are initiated only through the unit's KMC. In addition to managing the unit's data, the KMC is responsible for managing the propagation of the unit's data to every other C2 unit. The replications are handled through Mediator(s) for the C2 unit (MDC) at every other C2 unit. The KMC and MDCs for a given C2 unit must have identical software. Any changes to a unit's KMC management software needs to be synchronized with software changes to its MDCs at every other C2 unit. Knowledge managers and mediators are discussed in detail in Section 2.

Our approach uses a weak data synchronization mechanism between nodes, thereby allowing decisions to be made and actions taken based on inconsistent data and inferences. It requires management of historical data, keeping track of actions that are based on the use of possibly inconsistent estimated data, and methods for propagating changes when consistency is later achieved. The synchronization method uses sequence numbers on messages, and KMC aliveness messages and replies by MDCs are used to inform the KMC as to the state of the KMC's database on each remote node.

The overall result is support for timely data that at any point in time may be inconsistent but will converge over time toward a consistency or degree of correctness with respect to the real world. This approach allows the commander and his staff to use the near-real-time data immediately in decisionmaking and execution while also providing support for tracking data dependencies in decisions based on hypothetical data and propagating later critical data changes to the relevant actors. If data are late or missing, they may be estimated or inferred, labeled as hypothetical, and used in the KMC decisionmaking process.

The use of historical data and total data replication could support a very robust C2 system. Essentially, a commander and his staff could operate through their KMC at the node they select, which by preference would most likely be their command post node. By using replication, a commander whose node is in trouble could choose to move his KMC (and decisionmaking) to another node.

Examples of Handling C2 Events That Support Situational Awareness

For purposes of illustration this report walks through, at a high conceptual level, four events necessary to support situational awareness. These events indicate the kinds of system activities required to pass information horizontally and vertically for near-real-time processing concurrently throughout the entire system. They are (1) showing the ripple effect when a platoon has a location change that exceeds a threshold; (2) demonstrating the generation of a division-level Situation Report (SITREP) when one of its brigade's SITREPs is missing and has to be estimated; (3) illustrating how a change in a platoon's ability to perform its OPORD would ripple upward to corps and the dissemination of changed OPORDs from corps to platoon; and (4) showing the propagation of an intelligence report from platoon upward (i.e., due to sighting and engaging the enemy) and from corps downward.

Issues for Future Exploration

We identified military and technical issues needing more attention in the future.

Military Issues for Future Exploration

1. Need to bring together Joint Force and Army current and future doctrine, operational requirements, and future visions for battle management to interact with and drive proposed technological solutions.
2. Need to better understand concept of carousels of data shown in TRADOC PAM 525-5 (see Figure 1 in Section 1).
3. Need to explore doctrine to allow units to change their actions based on current situational assessment without or before receiving orders from their commanding unit.
4. Need to explore the effect of voice communications: If critical information is exchanged only by voice and not entered into the data system, then inferencing techniques including simulation will be limited.
5. Need to address data security issues including (a) risk at battalion and below, (b) dealing with classified data that have been downgraded to protect source; and (c) data aggregation risk posed by extensive replication.

Technical Issues for Future Exploration

1. Explore extensions to Object Oriented Database Management Systems (OODBMS) technology to support application-defined relationships and historical data.
2. Develop research concepts for use of long transactions and complex transactions to support the warfighter-decisionmaker.
3. Research use of knowledge management techniques for synchronization of data through propagation of critical changes.
4. Explore use of simulation to estimate behavior of other C2 nodes.
5. Explore categorization of types of data.
6. Investigate issues in use of DoD data standards based on relational technology with respect to moving to OODBMS technology.
7. Further explore robustness.

Download the Full Report ⤴