A Technical History of the SEI

general-purpose techniques but with special care to ensure that real-time constraints are satisfied.
For applications such as control systems and autonomous (air, land, or undersea) vehicles, as well
as many civilian applications (including medical devices and automobiles), real-time constraints,
while still important, are less a primary concern and are, in practice, treated more like other quality
attributes. The SEI pursued several efforts that provide engineering analysis to support treating
the real-time component as a quality-of-service (QoS) attribute.
Recognizing this trend, the SEI developed a suite of performance reasoning frameworks founded
on the principles of generalized rate monotonic analysis (GRMA) for predicting the average and
worst-case latency of periodic and stochastic tasks in real-time systems (Lambda-*). The Lambda-*
suite can be applied to many different, uniprocessor, real-time systems having a mix of tasks with
hard and soft deadlines with periodic and stochastic event inter-arrivals. Some examples include
embedded control systems (such as avionics, automotive, and robotic) and multimedia systems
(such as audio mixing). Tools were developed to check that a component-based design satisfied
various rules imposed by the reasoning framework. This enables the automatic generation of a
complete implementation of the design that would exhibit the runtime behavior “predicted” by the
reasoning framework, within an explicitly defined confidence interval. The important contribution
is that a user would only be able to design or build systems that exhibit predictable behavior by
construction, analogous to the way modern programming languages ensure that programs exhibit
memory safety by (type system) construction.
The SEI also pursued integrated methods for predictive analytic composition and tradeoff
(IMPACT) as a joint effort with CMU faculty and Lockheed Martin Aeronautics Company (LM-
Aero). The goal was the development of analytic methods to support the correct temporal composition
of systems. The methodology focused on the development of techniques to construct systems
having predictable timing performance and composed of pre-analyzed components. Resulting
methods included predictable dynamic assembly of software systems from pre-analyzed
“software parts” (PAAC), development of temporal analytic composition theory (TACT), predictive
models to utilize software and system-level performance measures, and engineering tradeoff
analyses involving both runtime attributes and design-time attributes [Saewong 2002].
These methods offered several benefits that support engineering tradeoff analyses at design time
and at runtime, including design uniformity using architectural patterns, reduction in rework
through system-level analysis conducted at design time, and the ability to address more complex
systems by leveraging pre-analysis of architectural patterns. These methods were used on the F-22
embedded avionics simulation to show that all temporal design characteristics expressed in the F-
22 challenge problem could be readily modeled and analyzed using a combination of real-time
queuing theory (RTQT) and generalized rate monotonic analysis techniques. Furthermore, it allowed
LM-Aero and CMU to propose a large-scale DASADA II experiment centered on upgrading
the F-22 mission computer. Results and insights from this experiment aimed to reduce both
new development and application rehost costs. The team was invited by the U.S. Army Aviation
and Missile Command (AMCOM) to propose a large-scale experiment centered on application of
technologies to the Sikorsky Black Hawk helicopter.
In the early 1990s, as recognition of the importance of software architecture grew, the SEI sought
ways to apply these emerging principles to real-time systems. A DARPA-funded effort that fostered
the creation of architecture description languages (ADLs) produced a design at Honeywell
CMU/SEI-2016-SR-027 | SOFTWARE ENGINEERING INSTITUTE | CARNEGIE MELLON UNIVERSITY 23
Distribution Statement A: Approved for Public Release; Distribution is Unlimited