Towards Adaptive & Reflective Middleware for Combat Systems Thursday, November 29, 2018 Executive Summary The Adaptive and Reflective Middleware Systems (ARMS) program focuses on mission-critical distributed and embedded systems where Different levels of service are possible and desirable under different conditions and costs; The levels of service in one dimension may need to be coordinated with and/or traded off against the levels of service in other dimensions to achieve the intended overall mission requirements; and Autonomous or semi-autonomous system behavior requires the networks, operating systems, and middleware components to be capable of reflection to adapt robustly to quantifiable changes in environmental conditions. In ARMS, middleware will be responsible for coordinating the exchange of information between remote entities efficiently, predictably, scalably, dependably, and securely by using advanced QoS capabilities of the underlying network and endsystems. In addition to automatically generating and optimizing middleware that supports the functional path, the ARMS program will focus on the QoS path that handles issues regarding how well the functional interactions behave end-to-end, e.g., the resources committed to client <=> object interaction, proper behavior when ideal resources are not available, the level of security needed, the recovery strategy for detected faults, etc. Key Themes of the Presentation Next-generation DoD systems will be substantially larger and more complex. e.g., they will be network-centric “systems of systems” that consist of many independent components that must collaborate in networks of embedded processors based increasingly on commercial-off-the-shelf (COTS) technology. Distributed object computing (DOC) middleware is increasingly essential to attain and sustain the success of such systems. DOC middleware helps to coordinate and control the COTS and non-COTS networking and computing resources to provide global end-to-end quality of service (QoS) properties. DARPA-funded R&D is needed to resolve the hard fundamental challenges related to DOC middleware that cannot be solved with today’s technology. The proposed research agenda is timely since industry and government are struggling to solve some of these challenges now. However, the DoD-related mission-critical properties of these challenges will not be solved by industry any time soon without significant DARPA investment. To support the DoD, an effective way forward for DARPA is to align and integrate research efforts with emerging standards-based COTS technology, which provides a realistic transition conduit. i.e., the DARPA R&D community must get “inside the COTS tornado” rather than trying to ignore or resist it. Historically, however, COTS introduced its own set of R&D challenges for mission-critical DoD distributed and embedded systems: e.g., inflexibility, lack of multiple integrated QoS properties, validation & verification, procurement , deployment & integration Key technology solutions that help resolve these COTS challenges include: Adaptive and reflective DOC middleware based on pattern languages and frameworks Leveraging and enhancing emerging standards in distributed and embedded real-time computing & communications High-confidence open-source software development processes and V&V techniques Dr. Douglas C. Schmidt dschmidt@darpa.mil DARPA/ITO Authorized for Public Release: Distribution Unlimited

Emerging Operational Trends & Challenges uav.navair.navy.mil/home.htm Emerging Trends Next-generation embedded systems are moving from platform-centric to network-centric distributed “systems of systems” Demands for greater autonomy are growing To meet expanding needs, our embedded systems must be more assurable, adaptable, & affordable Key Challenges Devising assurable embedded systems e.g., efficient, predictable, & safe, secure Devising adaptable embedded systems e.g., real-time dynamic allocation of computing/networking resources across many physical/virtual assets Devising affordable embedded systems e.g., transition technology to COTS

Emerging R&D Challenges & Trends Despite IT commoditization, COTS is often not applicable for mission-critical DoD network-centric embedded systems Middleware, Frameworks, & Components Patterns & Pattern Languages Standards & Open-source However, recent COTS software technology advances are helping to fundamentally reshape R&D High-performance, real-time, fault-tolerant, and secure systems Autonomous distributed embedded systems Key Themes of this Slide Despite the rapid commoditization of COTS information technology, there are many distributed and embedded DoD systems with mission-critical requirements that are hard to meet with today’s COTS technologies. Three general types of DoD systems that have this property are: Systems that run in the context of Navy Theater Wide (NTW) environments, where there’s a need for simultaneous support of multiple QoS properties, such as low-latency/jitter, high throughput, fault tolerance, and security. Autonomous distributed embedded systems, such as unmanned combat air vehicles (UCAVs) that coordinate in swarming missions. Power-aware, ad hoc, and mobile/wireless systems, such as those used to support small unit operations. To ensure long-term system affordability and make it easier to integrate future COTS hardware and software innovations, it’s important to improve COTS technologies so that they can be used in the mission-critical distributed and embedded systems outlined above. Although most COTS technologies are not sufficiently “ruggedized” to meet the needs of mission-critical DoD systems, there have been some important innovations in COTS software practice that provide opportunities to greatly enhance the development of DoD systems. Chief among these innovations include: Middleware consisting of pluggable service & micro-protocol components & reusable “semi-complete” application frameworks Patterns and pattern languages, which help generate software architectures by capturing recurring structures & dynamics & by resolving design forces New standards (such as Real-time CORBA and Real-time Java) and new development processes, such as pen-source, refactoring, extreme programming (XP), and advanced V&V techniques Power-aware ad hoc, mobile, distributed, & embedded systems

Embedded System Approaches Problems with Current Embedded System Approaches Dynamic embedded system QoS requirements historically not supported by COTS i.e., COTS is too big, slow, buggy, incapable, & inflexible Likewise, the proprietary multiple technology bases in embedded systems today limit effectiveness by impeding Assurability (of QoS), Adaptability, & Affordability Today, each combat system brings its own: networks computers displays software people Applications Applications Problems Non-scalable tactical performance Inadequate QoS control for joint operations e.g., distributed weapons control High software lifecycle costs e.g., many “accidental complexities” & low-level platform dependencies Sensor Systems Command & Control System Engagement System Weapon Control Systems Weapon Systems EO Kill Eval Sched Illum Network AAW EG TBM EG AAW AAW MG AAW AAW TMB MG AAW Technology base: Proprietary MW Mercury Link16/11/4 Technology base: DII-COE POSIX ATM/Ethernet Technology base: Proprietary MW POSIX NTDS Technology base: Proprietary MW VxWorks FDDI/LANS Technology base: Proprietary MW POSIX VME/1553 Operating System Operating System Endsystem Wireless/Wireline Networks Endsystem

A More Effective Approach Create the new generation of adaptive & reflective middleware system (ARMS) technologies to simultaneously control multiple system QoS properties Adaptive – capable of static or dynamic modification Reflective – capable of self-adaptation based on functional & QoS context QoS – non-functional system properties, e.g., thruput, latency/jitter, scalability, dependability, & security Applications Applications Sensor System Weapon Command & Control System Engagement Weapon Control Middleware } ARMS Benefits Highly scalable tactical performance e.g., distributed resource mgmt. Enable new warfighting capability e.g., distributed weapons control Support common technology bases e.g., elevate standardization of COTS to middleware to control software lifecycle costs by minimizing lower-level dependencies Domain-Specific Services } Common Services Distribution Middleware Infrastructure Middleware Operating System Wireless/Wireline Networks Endsystem Endsystem

DARPA/ITO Family of Embedded Systems Programs MoBIES Design technology & software CAD Hardware Domain-Specific Services Common Distribution Middleware Infrastructure Operating Systems & Protocols Applications SEC Hybrid system control & computation ARMS Adaptive & reflective middleware Quorum Quality-of-service & translucent layers PCES Composable embedded systems NEST Deeply networked embedded systems PCA Polymorphous computing architecture

Why We Can Make a Difference Now Recent synergistic advances in fundamentals: QoS-enabled Middleware Pluggable protocol/service components & reusable “semi-complete” frameworks that assure end-to-end system qualities Distributed Resource Modeling Distributed Resource Modeling Formally specify resource mgmt. plans; model, reason about, & refine them; & monitor/enforce them automatically QoS-enabled Middleware Pattern Languages Pattern Languages Generate software architectures by (1) capturing common structures & dynamics & (2) resolving design constraints Recent success infusing standards-based COTS into DRE systems: DRTS Java RT Linux RT CORBA ARMS WSOA Real-time Retargeting hi DRE researchers Key themes Middleware is becoming increasingly QoS-enabled e.g., the Real-time CORBA 1.0 specification and the RT Java specification Patterns represent successful solutions to problems that arise when building software. When related patterns are woven together, they form a language that helps to Define a vocabulary for talking about software development problems; and Provide a process for the orderly resolution of these problems. Studying and applying patterns and pattern languages helps developers enhance software quality by addressing fundamental challenges in large-scale system development. These challenges include communication of architectural knowledge among developers; accommodating new design paradigms or architectural styles; resolving non-functional forces, such as reusability, portability, and extensibility; and avoiding development traps and pitfalls that are usually learned only by costly trial and error. Middleware frameworks/components and pattern languages address inventions, whereas open-source addresses dissemination. DRE practitioners C++ UNIX CORBA Java Linux RT CORBA Level of DRE Tactical Technology Abstraction Bold Stroke Avionics Mission Computing Network C/Ada Cyclic execs Proprietary C++ UNIX CORBA RT Java RT Linux RT CORBA lo ’90-’95 ’96-’01 ’02-’06