Investigating the Tradespace between Increased Automation and Optimal Manning on Aircraft Carrier Decks

With increasing prevalence and capabilities of autonomous systems as part of complex heterogeneous manned-unmanned environments (HMUEs), an important consideration is the impact of the introduction of automation on the optimal assignment of human personnel. The US Navy has implemented optimal staffing techniques before in the 1990's and 2000's with a "minimal staffing" approach. The results were poor, leading to the degradation of Naval preparedness. Clearly, another approach to determining optimal staffing is necessary. To this end, the goal of this research is to develop human performance models for use in determining optimal manning of HMUEs. The human performance models are developed using an agent-based simulation of the aircraft carrier flight deck, a representative safety-critical HMUE. The Personnel Multi-Agent Safety and Control Simulation (PMASCS) simulates and analyzes the effects of introducing generalized maintenance crew skill sets and accelerated failure repair times on the overall performance and safety of the carrier flight deck. A behavioral model of four operator types (ordnance officers, chocks and chains, fueling officers, plane captains, and maintenance operators) is presented here along with an aircraft failure model. The main focus of this work is on the maintenance operators and aircraft failure modeling, since they have a direct impact on total launch time, a primary metric for carrier deck performance. With PMASCS I explore the effects of two variables on total launch time of 22 aircraft: 1) skill level of maintenance operators and 2) aircraft failure repair times while on the catapult (referred to as Phase 4 repair times). It is found that neither introducing a generic skill set to maintenance crews nor introducing a technology to accelerate Phase 4 aircraft repair times improves the average total launch time of 22 aircraft. An optimal manning level of 3 maintenance crews is found under all conditions, the point at which any additional maintenance crews does not reduce the total launch time. An additional discussion is included about how these results change if the operations are relieved of the bottleneck of installing the holdback bar at launch time.

Studying Recommender Systems to Enhance Distributed Computing Schedulers

Distributed Computing frameworks belong to a class of programming models that allow developers to

launch workloads on large clusters of machines. Due to the dramatic increase in the volume of

data gathered by ubiquitous computing devices, data analytic workloads have become a common

case among distributed computing applications, making Data Science an entire field of

Computer Science. We argue that Data Scientist's concern lays in three main components: a dataset,

a sequence of operations they wish to apply on this dataset, and some constraint they may have

related to their work (performances, QoS, budget, etc). However, it is actually extremely

difficult, without domain expertise, to perform data science. One need to select the right amount

and type of resources, pick up a framework, and configure it. Also, users are often running their

application in shared environments, ruled by schedulers expecting them to specify precisely their resource

needs. Inherent to the distributed and concurrent nature of the cited frameworks, monitoring and

profiling are hard, high dimensional problems that block users from making the right

configuration choices and determining the right amount of resources they need. Paradoxically, the

system is gathering a large amount of monitoring data at runtime, which remains unused.

In the ideal abstraction we envision for data scientists, the system is adaptive, able to exploit

monitoring data to learn about workloads, and process user requests into a tailored execution

context. In this work, we study different techniques that have been used to make steps toward

such system awareness, and explore a new way to do so by implementing machine learning

techniques to recommend a specific subset of system configurations for Apache Spark applications.

Furthermore, we present an in depth study of Apache Spark executors configuration, which highlight

the complexity in choosing the best one for a given workload.