Advances in technologies like artificial intelligence and metaverse have led to a proliferation of software systems in business and everyday life. With this widespread penetration, the carbon emissions of software are rapidly growing as well, thereby negatively impacting the long-term sustainability of our environment. Hence, optimizing software from a sustainability standpoint becomes more crucial than ever. We believe that the adoption of automated tools that can identify energy-inefficient patterns in the code and guide appropriate refactoring can significantly assist in this optimization. In this extended abstract, we present an industry case study that evaluates the sustainability impact of refactoring energy-inefficient code patterns identified by automated software sustainability assessment tools for a large application. Preliminary results highlight a positive impact on the application’s sustainability post-refactoring, leading to a 29% decrease in energy consumption.
An Energy-Aware Approach to Design Self-Adaptive AI-based Applications on the Edge
The advent of edge devices dedicated to machine learning tasks enabled the execution of AI-based applications that efficiently process and classify the data acquired by the resource-constrained devices populating the Internet of Things. The proliferation of such applications (e.g., critical monitoring in smart cities) demands new strategies to make these systems also sustainable from an energetic point of view.
In this paper, we present an energy-aware approach for the design and deployment of self-adaptive AI-based applications that can balance application objectives (e.g., accuracy in object detection and frames processing rate) with energy consumption. We address the problem of determining the set of configurations that can be used to self-adapt the system with a meta-heuristic search procedure that only needs a small number of empirical samples. The final set of configurations are selected using weighted gray relational analysis, and mapped to the operation modes of the self-adaptive application.
We validate our approach on an AI-based application for pedestrian detection. Results show that our self-adaptive application can outperform non-adaptive baseline configurations by saving up to 81% of energy while loosing only between 2% and 6% in accuracy.
SmartCoCo: Checking Comment-code Inconsistency in Smart Contracts via Constraint Propagation and Binding
Smart contracts are programs running on the blockchain. Comments in source code provide meaningful information for developers to facilitate code writing and understanding. Given various kinds of token standards in smart contracts (e.g., ERC-20, ERC-721), developers often copy&paste code from other projects as templates, and then implement their own logic as add-ons to such templates. In many cases, the consistency between code and comment is not well-aligned, leading to comment-code inconsistencies (as we call CCIs). Such inconsistencies can mislead developers and users, and even introduce vulnerabilities to the contracts. In this paper, we present SmartCoCo, a novel framework to detect comment-code inconsistencies in smart contracts. In particular, our research focuses on comments related to roles, parameters, and events that may lead to security implications. To achieve this, SmartCoCo takes the original smart contract source code as input and automatically analyzes the comment and code to find potential inconsistencies. SmartCoCo associates comment constraints and code facts via a set of propagation and binding strategies, allowing it to effectively discover inconsistencies with more contextual information. We evaluated SmartCoCo on 101,780 unique smart contracts on Ethereum. The evaluation result shows that SmartCoCo achieves good effectiveness and efficiency. In particular, SmartCoCo reports 4,732 inconsistencies from 1,745 smart contracts, with a precision of over 79% on 439 manual-labeled comment-code inconsistencies. Meanwhile, it only takes 2.64 seconds to check a smart contract on average.
With the exponential growth of data, the demand for effective data analysis tools has increased significantly. R language, known for its statistical modeling and data analysis capabilities, has become one of the most popular programming languages among data scientists and researchers. As the importance of energy-aware software systems continues to rise, several studies investigate the impact of source code and different stages of machine learning model training on energy consumption. However, existing studies in this domain primarily focus on programming languages like Python and Java, leaving a gap for energy measuring tools in other programming language such as R. To address this gap, we propose ``\textbf{\textit{RJoules}}'', a tool designed to measure the energy consumption of R code snippets. We evaluate the correctness and performance of \textit{RJoules} by applying it to four machine learning algorithms on three different systems. Our aim is to support developers and practitioners in building energy-aware systems in R. The demonstration of the tool is available at \url{https://youtu.be/yMKFuvAM-DE} and related artifacts at https://rishalab.github.io/RJoules.
Assessing the Impact of Refactoring Energy-Inefficient Code Patterns on Software Sustainability: An Industry Case Study
Recorded talk
Advances in technologies like artificial intelligence and metaverse have led to a proliferation of software systems in business and everyday life. With this widespread penetration, the carbon emissions of software are rapidly growing as well, thereby negatively impacting the long-term sustainability of our environment. Hence, optimizing software from a sustainability standpoint becomes more crucial than ever. We believe that the adoption of automated tools that can identify energy-inefficient patterns in the code and guide appropriate refactoring can significantly assist in this optimization. In this extended abstract, we present an industry case study that evaluates the sustainability impact of refactoring energy-inefficient code patterns identified by automated software sustainability assessment tools for a large application. Preliminary results highlight a positive impact on the application’s sustainability post-refactoring, leading to a 29% decrease in energy consumption.
Green AI Quotient : Assessing Greenness of AI-based software and the way forward
Recorded talk
As the world takes cognizance of AI’s growing role in greenhouse gas(GHG) and carbon emissions, the focus of AI research & development is shifting towards inclusion of energy efficiency as another core metric. Sustainability, a core agenda for most organizations, is also being viewed as a core nonfunctional requirement in software engineering. A similar effort is being undertaken to extend sustainability principles to AI-based systems with focus on energy efficient training and inference techniques. But an important question arises, does there even exist any metrics or methods which can quantify adoption of “green” practices in the life cycle of AI-based systems? There is a huge gap which exists between the growing research corpus related to sustainable practices in AI research and its adoption at an industry scale. The goal of this work is to introduce a methodology and novel metric for assessing ”greenness” of any AI-based system and its development process, based on energy efficient AI research and practices. The novel metric, termed as Green AI Quotient, would be a key step towards AI practitioner’s Green AI journey. Empirical validation of our approach suggest that Green AI Quotient is able to encourage adoption and raise awareness regarding sustainable practices in AI lifecycle.
