Sowmya Karunakaran

Bio: Sowmya Karunakaran is an academic researcher from Google. The author has contributed to research in topics: Cloud computing & Bidding. The author has an hindex of 6, co-authored 12 publications receiving 94 citations. Previous affiliations of Sowmya Karunakaran include Indian Institutes of Technology & Indian Institute of Technology Madras.

Bidding Strategies for Spot Instances in Cloud Computing Markets

Abstract: In recent times, spot pricing -- a dynamic pricing scheme -- is becoming increasingly popular for cloud services. This new pricing format, though efficient in terms of cost and resource use, has added to the complexity of decision making for typical cloud computing users. To recommend bidding strategies in spot markets, we use a simulation study to understand the implications that provider-recommended strategies have for cloud users. We use data based on Amazon's Elastic Compute Cloud spot market to provide users with guidelines when considering tradeoffs between cost, wait time, and interruption rates.

Data Breaches: User Comprehension, Expectations, and Concerns with Handling Exposed Data

Abstract: Data exposed by breaches persist as a security and privacy threat for Internet users. Despite this, best practices for how companies should respond to breaches, or how to responsibly handle data after it is leaked, have yet to be identified. We bring users into this discussion through two surveys. In the first, we examine the comprehension of 551 participants on the risks of data breaches and their sentiment towards potential remediation steps. In the second survey, we ask 10,212 participants to rate their level of comfort towards eight different scenarios that capture real-world examples of security practitioners, researchers, journalists, and commercial entities investigating leaked data. Our findings indicate that users readily understand the risk of data breaches and have consistent expectations for technical and non-technical remediation steps. We also find that participants are comfortable with applications that examine leaked data—such as threat sharing or a “hacked or not” service—when the application has a direct, tangible security benefit. Our findings help to inform a broader discussion on responsible uses of data exposed by breaches.

Business view of cloud

Abstract: Purpose – This study aims to investigate the decisions related to business aspects of cloud computing and discuss the research density, models/techniques used and identify opportunities for future work. Design/methodology/approach – In this paper, 155 research articles shortlisted through a systematic review were analyzed and a classification framework was developed. Using this framework, the research density is discussed and a detailed review of four widely researched decision themes is provided. Findings – It was found that current research on business aspects is spread across 23 decision themes. The distribution, however, is skewed with 50 per cent pertaining to just four themes, namely, pricing, markets, sourcing and adoption. Simulation appears to be the preferred modeling approach. Decision themes in consumer behavior, sustainability, auditing and culture offer opportunities for future research. Research limitations/implications – The classification framework organizes extant research on applied mod...

Testing Stylistic Interventions to Reduce Emotional Impact in Content Moderation Workers

Abstract: With the rise in user generated content, there is a greater need for content reviews. While machines and technology play a critical role in content moderation, the need for manual reviews still remains. It is known that such manual reviews could be emotionally challenging. We test the effects of simple interventions like grayscaling and blurring to reduce the emotional impact of such reviews. We demonstrate this by bringing in interventions in a live content review setup thus allowing us to maximize external validity. We use a pre-test post-test experiment design and measure review quality, average handling time and emotional affect using the PANAS scale. We find that simple grayscale transformations can provide an easy to implement and use solution that can significantly change the emotional impact of content reviews. We observe, however, that a full blur intervention can be challenging to reviewers.

Decisions, Models and Opportunities in Cloud Computing Economics: A Review of Research on Pricing and Markets

Abstract: Cloud computing has emerged as a key information technology and systems model over the last few years. Major organizations have developed and delivered cloud computing solutions and continue to do so. Consequently, a number of strides were made in the advancement of technology leading to a growth in the adoption of cloud computing. The growing recognition of cloud computing services necessitates a focus on the business aspects of cloud. However, we feel that research in this area is scant. To this effect, we performed a systematic review of cloud computing literature and reviewed 2891 abstracts and 157 articles published until the year 2012. Based on the findings of the review, we establish a framework for organising the extant research on cloud business aspects. Using the framework, we find cloud economics to be the most widely researched business aspect. In this paper, we provide a detailed review of the application of decision models in the context of cloud economics, with a specific focus on pricing and markets. The proposed framework and review results serve as a reference to IS researchers and practitioners to understand decision situations, models and opportunities.

Amazon Web Services完全ソリューションガイド

Abstract: １ インフラを構築する（ＡＷＳにおけるインフラ；ＶＰＣを構成する；ＶＰＣとオンプレミス環境とを接続する） ２ ファイルオブジェクトを保存・共有・公開する（オブジェクトストレージＳ３の機能；ファイルストレージとして利用する；Ｗｅｂサーバーを構築する；信頼性とコストのバランスをとりたい） ３ アプリケーションサーバーを構築する（Ａｍａｚｏｎ ＥＣ２とＡＷＳ Ｌａｍｂｄａ；スケーラビリティーを高める；サーバーレスでプログラムを動かす；データベースサービスを活用する） ４ ＡＷＳシステムを管理する（リソース監視と異常検知・通報；耐障害性を高める仕組みとバックアップ＆リカバリー；構成管理）

Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned

Abstract: We describe our early efforts to red team language models in order to simultaneously discover, measure, and attempt to reduce their potentially harmful outputs. We make three main contributions. First, we investigate scaling behaviors for red teaming across 3 model sizes (2.7B, 13B, and 52B parameters) and 4 model types: a plain language model (LM); an LM prompted to be helpful, honest, and harmless; an LM with rejection sampling; and a model trained to be helpful and harmless using reinforcement learning from human feedback (RLHF). We find that the RLHF models are increasingly difficult to red team as they scale, and we find a flat trend with scale for the other model types. Second, we release our dataset of 38,961 red team attacks for others to analyze and learn from. We provide our own analysis of the data and find a variety of harmful outputs, which range from offensive language to more subtly harmful non-violent unethical outputs. Third, we exhaustively describe our instructions, processes, statistical methodologies, and uncertainty about red teaming. We hope that this transparency accelerates our ability to work together as a community in order to develop shared norms, practices, and technical standards for how to red team language models.

The Psychological Well-Being of Content Moderators: The Emotional Labor of Commercial Moderation and Avenues for Improving Support

Abstract: An estimated 100,000 people work today as commercial content moderators. These moderators are often exposed to disturbing content, which can lead to lasting psychological and emotional distress. This literature review investigates moderators’ psychological symptomatology, drawing on other occupations involving trauma exposure to further guide understanding of both symptoms and support mechanisms. We then introduce wellness interventions and review both programmatic and technological approaches to improving wellness. Additionally, we review methods for evaluating intervention efficacy. Finally, we recommend best practices and important directions for future research. Content Warning: we discuss the intense labor and psychological effects of CCM, including graphic descriptions of mental distress and illness.