Program managers need systems that can perform successfully under adverse circumstances—for example, under heavy loads or in the presence of subsystem failures. Yet the behavior of systems under such circumstances is often unacceptable. The critical need to manage performance is obvious in real-time systems, such as flight-control software. Likewise, unexpected performance problems in command-and-control or management information systems can make such systems virtually unusable until costly repairs are undertaken.

Purpose

The purpose of SEI work in performance-critical systems is to help ensure that both the government and its contractors are aware of effective techniques for predicting and controlling critical aspects of system performance. In addition, because new techniques are evolving for controlling critical system-performance properties, the SEI is also committed to bringing the best of emerging practices into use on DoD systems.

Equally important are techniques for ensuring the robustness of systems even in the presence of software errors, unexpected patterns of use, and hardware failures.

In the coming year, the SEI will refocus this work on technology to address the DoD’s increasing needs for highly dependable distributed and networked systems, as well as on survivability. The accomplishments specified in the next section reflect the focus in FY2002 on performance.

2002 Accomplishments

Portable Operating System Interface Standard
(POSIX) Supported
The SEI has been a key agent for the government in obtaining a standard for real-time distributed systems communication. An SEI staff member chaired the POSIX working group of the Institute of Electrical and Electronics Engineers (IEEE): Real-Time Distributed Systems Communications, Language Independent Specification. SEI staff members helped to develop and voted on the standard. Reballoting was completed in FY2002, and the standard was approved. POSIX 1003.21 will bring standardization for the UNIX community where none had previously existed and thus provide the potential for increased portability of software.

Navy DD(X) Program Supported
The SEI helped the Navy’s DD(X) Program in the areas of system and software architecture by participating in technical reviews for a variety of DD(X) subsystems, including the multi-function radar (MFR) subsystem, and providing recommendations to the contractor and the program office. In addition, the SEI developed and presented a well-attended half-day tutorial for the entire DD(X) Program Office on performance-critical systems and software architecture. The SEI has been actively participating on the system architecture, software engineering, and executive review panels for the DD(X) Program Office Total Ship Computing Environment technical team.

Avionics Architecture Description Language Developed
The SEI is helping develop a standard for an Avionics Architecture Description Language (AADL), under the auspices of the Society of Automotive Engineers (SAE) Avionic Systems Division (ASD) and with funding from U.S. Army Aviation and Missile Command (AMCOM). The standard provides descriptive methods that permit the formal analysis of system-performance properties. The SEI has been an evaluator and user of the technology that underlies the standard, and a staff member is co-author and editor of the standards document.

The establishment of the AADL is having widespread impact. The standard itself is being aligned with the Unified Modeling Language (UML) to ensure an active practitioner community. Tutorials have been offered on AADL at SAE meetings and at the Digital Avionics Systems Conference. A recent tutorial and coordination meeting in Toulouse, France, was attended by 90 people from the avionics, aerospace, and automotive industries. This community is starting to investigate the incorporation of AADL into its real-time software engineering practices.

Future Combat Systems Program Supported
The SEI has been actively supporting the analysis of software issues for the Army’s Future Combat Systems (FCS) program. The FCS program is in its early phase of concept and technology development, when improved software-development methods can have maximum impact. Initial activities have included evaluation of the software-development plan and the proposed FCS architecture.

This effort involved collaboration with the Center for Empirically Based Software Engineering (CeBASE). Principal collaborators were drawn from the University of Maryland, the Fraunhofer Center-MD, and the University of Southern California Center for Software Engineering.