Environmental and Occupational Health Sciences Institute
Rutgers University Community RepositoryNot a member yet
58345 research outputs found
Sort by
Toward scalable and high-performance I/O with cross-layered storage design
No full textThe storage technologies have been rocketing in the past decade, PM (persistent memory) offers nano-second scale latency while NVMe SSD provide bandwidth of tens of gigabytes. However, software innovations in managing storage devices fall behind such hardware advancement. Firstly, traditional I/O stack and file systems in operating systems, widely used in both industry and academia, suffer from high software overheads (e.g., system calls, data copies, coarse-grained locks), which are no longer negligible for emerging fast storage devices with micro-second hardware latency. Secondly, with the increasing compute power and memory within storage devices, near-data processing provides opportunities for redesigning software I/O stacks and interfaces to keep pace with fast storage hardware. However, traditional I/O stack in operating system and existing user-level solutions are not able to exploit such opportunities without an efficient I/O interface to bind I/O requests to their related data processing operations. Thirdly, exa-scale data with explosive growth motivates the collaborative use of heterogeneous storage devices with both emerging and existing storage media. Nevertheless, managing these devices with vast differences in bandwidth, asymmetrical read and write performance, and capacity gaps present significant challenges.
This thesis addresses these problems with a cross-layered system design by dividing I/O software stack into user space, OS kernel, and storage firmware to capitalize on the advantages of each layer. The research scope in this thesis diverges in two dimensions. The first is to redesign I/O stacks and interfaces at a single storage level. The second is to develop methods of managing heterogeneous storage devices.
At a single storage level, to reduce software overheads in traditional I/O stack in operating systems with emerging computational storage devices, the first work in this thesis proposes a cross-layered file system design that disaggregates file systems into user-space, OS, and storage devices to eliminate scalability bottlenecks in each layer and scaling I/O performance with fine-grained concurrency. We demonstrate the file descriptor instead of the inode should server as the abstraction to concurrently access files. By mapping file descriptors to dedicated hardware I/O queues, we efficientlyeliminate coarse-grained locks used in existing systems.
On top of the first work, this thesis made the observation that there are computational tasks tightly coupled with I/O operations, causing unnecessary context switching and data copy overheads potentially. The second work proposes a novel CISC I/O interface compatible with POSIX which packs multiple I/O operations with related computational tasks in a single compound I/O operation to utilize the compute power in storage devices and reduces data copy overheads effectively.
Scaling beyond the scope of a single storage, we target heterogeneous storage systems composed of different types of storage devices. Existing caching and tiering approaches for managing heterogeneous storage devices place a faster storage on top of the storage hierarchy, which fails to fully capitalize on the cumulative storage capabilities. The third work in this thesis presents a novel solution to exploit the collective hardware and software capabilities offered by multiple storage devices by delegating resource management to user space and retaining important properties such as permission enforcement and sharing in the OS kernel.
In summary, motivated by the software bottlenecks in existing solutions, this thesis proposes a cross-layered design to disaggregate the components of the I/O stack across user space, OS kernel, and storage firmware (if applicable) to reduce the software overheads incurred from a monolithic I/O stack design in most existing solutions.Ph.D.Includes bibliographical reference
Development and validation of analytical strategies for the measurement of migratory isocyanates and primary amines from food packaging lamination adhesives
No full textLaminated food packaging materials containing polyurethane lamination adhesives are used by industry to impart the functional properties of multiple materials into a single structure to enhance packaging performance. Polyurethane adhesives are made by reaction of poly-isocyanates with polyols, forming a high molecular weight adhesive network. Polyurethanes must be cured under controlled conditions before use to ensure that the reactive isocyanates are completely consumed in the reaction. If inadequately cured, residual isocyanates in the adhesive can potentially migrate into foods or react with atmospheric moisture, forming primary aromatic amines. Isocyanates and primary aromatic amines are considered potently carcinogenic, and regulatory agencies mandate extremely low detection limits for these compounds in food-contact materials.
The goal of this research project was to develop and validate novel analytical strategies for measuring these compounds in food packaging matrices. For each class of target compound, strategies using different analytical techniques or instruments were developed to obtain maximum sensitivity as well as to demonstrate analytical options available to scientists.
Migration testing using 95% ethanol food simulant was used to simultaneously extract isocyanates and quantitatively convert them into ethylurethanes. High performance liquid chromatography-high resolution accurate mass-mass spectrometry (HPLC-HRAM-MS) and gas chromatography-mass spectrometry (GC-MS) assays were explored; validated detection limits as low as 0.05 ppb w/v were achieved for isocyanate-(ethylurethanes) using HPLC-HRAM-MS.
Migration testing using 10% ethanol food simulant was used to extract primary amines from packaging samples. Extracts were subsequently analyzed by HPLC-HRAM-MS or derivatized using N-(heptafluorobutyryl)imidazole (HFBI) and analyzed by GC-MS. Validated detection limits for primary amines were as low as 0.1 ppb w/v using HPLC-HRAM-MS or GC-MS.
Eleven developmental laminated packaging materials were subjected to the developed migration testing and analysis assays. Migratory isocyanates and primary amines were detected in several test samples at sub-part-per-billion levels. Comparison of assays showed that migratory concentrations between techniques were generally in agreement, demonstrating validity of the developed methods. The benefits and shortcomings of each assay were thoroughly discussed.
This present research will serve as a reference document to demonstrate highly sensitive and selective analytical methods for the analysis of migratory isocyanates and primary amines from laminated food packaging materials.Ph.D.Includes bibliographical reference
Construction, analysis, and verification of a 3D-printed carbon fiber-silicone composite system for vibration isolation
No full textEmbarking on a thorough investigation of vibration isolation, we employ an innovative 3D-printed carbon fiber-silicone composite system. The material modeling phase, featuring the fabrication of a 3D print carbon fiber-silicone composite cube, selected viscoelastic model, lays the groundwork for dynamic mechanical analysis (DMA) experiments conducted on a 10x10x10mm composite cube. Key insights, including heightened stiffness and improved damping characteristics, foundational for effective vibration isolation, are revealed through these experiments. The subsequent integration of Numerical Analysis validates DMA findings. Data Analysis and Modelling constructs a robust look-up table for material properties, enhancing understanding of the composite's behavior under dynamic conditions.
In the empirical realm, free displacement tests are performed on aluminum and carbon fiber-silicone strips, affirming the efficacy of the composite in reducing vibrations. This tangible evidence positions the composite as a practical solution for vibration isolation challenges. This study proposes a forward-looking vibration displacement experiment to assess real-world effectiveness beyond laboratory confines. This interdisciplinary research, emphasizing passive voice throughout, contributes foundational knowledge on composite materials, portraying the 3D-printed carbon fiber-silicone composite as a promising tool for engineers and industries grappling with vibration-related challenges. The study marks a significant step towards redefining vibration isolation technologies, with the 3D-printed composite system emerging as a potential revolution in the domain of advanced materials for vibration isolation applications.
In conclusion, the research not only delves into the intricate dynamics of vibrations but also positions the study as an opportunity to redefine engineering solutions. As this journey unfolds, the 3D-printed composite system emerges as a potential paradigm shift in advanced materials, presenting a unique proposition for vibration isolation applications. Beyond immediate applications in vibration control, the study aligns with broader pursuits of sustainable technologies, minimizing energy consumption, reducing environmental impact, and offering versatile solutions to multifaceted challenges.M.S.Includes bibliographical reference
Pediatric surgery quality improvement: decreasing rates of perforated appendicitis and post-operative abscess formation
No full textPurpose of Project: The aim of this quality improvement project was to standardize care of children with appendicitis and perforated appendicitis using evidence-based protocols at two urban medical centers in New Jersey. Additionally, socioeconomic factors contributing to patients presenting with perforated appendicitis were explored.
Methods: A retrospective chart review of data collected as part of routine care was conducted. Chi-square tests were used to determine differences in practice prior to and after implementation of the clinical management protocols. Perforation rates from the data collection were analyzed with regional socioeconomic data from US Census Data.
Results: Prescribing of ceftriaxone and metronidazole as first line treatment for appendicitis increased after implementation of the standardized protocol (13 cases versus 40; p=<.001). Adherence with the clinical management protocol was higher for patients treated at one of the project sites compared to patients transferred from another facility (40 cases versus 13; p=.023). Additionally, adherence was greater during day shift hours (7a-7p, n=30) versus night shift hours (7p-7a, n=10; p=.001). While there was no change in imaging modality (ultrasound vs CT), CT scans were ordered more frequently in patients transferred from an outside hospital compared to those seen at the project sites (53 cases versus 40; p=<.001). Factors associated with lower likelihood of undergoing an ultrasound included patients without private insurance (p=<.001), those without a pediatrician (p=.047), and those whose parents did not speak English as the primary language (p=.035). There was a high rate of perforation in patients with non-English speaking parents (p=.019).
Implications: Use of a standardized clinical management protocol for appendicitis improves adherence with antibiotic selection and reduces variability in care and improves outcomes. Reasons for non-compliance of protocols must be examined.D.N.P.Includes bibliographical reference
Mind the Gap: improving post-operative dilation care in gender affirming vaginoplasty
No full textPurpose of Project: This Doctor of Nursing Practice (DNP) quality improvement (QI) project aimed to address this gap by developing a consensus-based multidisciplinary dilation toolkit protocol for transgender, gender non-binary, and intersex (TGNBI) individuals. The aim was to improve outcomes, enhance the patient experience, and optimize care.
Methods: The objectives were to create a dilation toolkit and protocol that experts in the field of gender affirming care and surgery could review and find consensus or not find consensus for use of the toolkit for clinical experts within a 12-week timeframe. Participants for this project were recruited through an appeal on the World Professional Association of Transgender Health (WPATH) member-only message board, as well as personal networking based on previous relationships. Using the Appraisal of Guidelines for Research & Evaluation II (AGREE II) instrumentation, comparisons were made between clinical roles and the methodological quality of the toolkit and whether they would recommend it for use or not.
Results: Although there was a statistically significant lower score for the rigor of the toolkit between certain clinical disciplines, the recommendation of the toolkit was unanimous. Therefore, consensus was attained.
Implications for Practice: Future research should be muti-faceted and examine outcomes with the dilation protocol and teaching, increased patient engagement, adherence to dilation, and reduced revision rates. Additionally, research should examine outcomes related to post dilation frequencies for more scientific standard that can be applied to all dilation. Moreover, a toolkit focused on patient perspectives, input, and guidance around usefulness should be considered.D.N.P.Includes bibliographical reference
Blue, green and grey: a concrete solution
No full textThis thesis examines the environmental impact of concrete and the materials that have emerged to ameliorate this impact. Of particular interest, is how a carbon intensive material like concrete can be sustainably used to create viable ecological habitats. A significant part of this inquiry involves the ecological potential of concrete to address various environmental issues, exemplified by its role in the creation of purpose made structures such as artificial reefs, living walls and floating wetlands. Transitioning from theoretical analysis to practical application, this thesis culminates in a design proposal tailored to the existing ecological conditions of Bellandur Lake, a 900-acre freshwater reservoir located in the south Indian city of Bangalore. The design proposal utilizes floating concrete vessels as a proactive solution to address the unique environmental challenges of the lake and the restoration of its ecosystem, thus providing a holistic perspective on the sustainable use of concrete, particularly in the context of environmental conservation and urban development.M.L.A.Includes bibliographical reference
The Robison Hebrew manuscript collection: 50 Yemenite treasures in our midst: 35th annual Louis Faugères Bishop III lecture
No full textRutgers Distinguished Professor Gary A. Rendsburg delivers the 35th annual Louis Faugères Bishop III Lecture, entitled “The Robison Hebrew Manuscript Collection: 50 Yemenite Treasures in Our Midst,” on Tuesday, April 2, at 4:00 p.m., at Alexander Library, New Brunswick, NJ.
The Bishop Lectures feature diverse topics on book and manuscript collecting, printing history, and the use of rare books and manuscripts. The series is named in memory of Louis Faugères Bishop Jr.’s son, a prominent cardiologist and book lover who helped build one of the excellent New York private libraries at the New York Racquet Club
Graph neural network and motion planning
No full textGraph neural networks (GNNs) have been widely used in various practical applications such as fake news detection, drug discovery, and content recommendation. Despite their success and popular adoptions, GNNs suffer the intentional attacks with imperceptible changes in data. To be specific, the attackers can perturb the node features or graph structure explicitly, thereby causing the wrong prediction of a specific target node. However, performing specific attacks on different targets can be expensive in real scenarios. In the first part of our thesis, we study the vulnerability of graph learning models and design the universal attack on node classification tasks. More specifically, we identify a set of a few bad nodes from the existing graph as the anchors, and any target node can be attacked by flipping its connections with the anchors. However, sometimes it is unrealistic to gain the control over the anchors nodes, therefore, we propose another method that adds fake nodes as the anchors and builds the edges between fake nodes and real nodes. Any node can be effectively attacked by the fake nodes without impacting the overall accuracy of other nodes, which is hard to detect.
The second part of this dissertation focuses on the study of motion planning, which is one of the fundamental tasks across many applications like drone navigation, autonomous driving and robot arm manipulation, etc. Neural network (NN)-based methods have emerged as an attractive approach for robot motion planning due to the strong learning capabilities of NN models and their inherently high parallelism. We formulate the motion planning as a video prediction problem so that the future robot movement can be predicted as video frames iteratively. However, describing the planning statuses using images is not effective in high-dimensional motion planning tasks such as KUKA 14D. Therefore, we formulate motion planning as graph search problems, apply GNNs to construct collision-free random geometric graph and predict high-quality heurisitic values to guid the A* search. Such GNN-based planner can search for the high-quality paths with very fast planning speed under any-dimensional tasks.Ph.D.Includes bibliographical reference
Rutgers-Camden Graduate Research and Creative Works Symposium 2024. Iyer, Anusha_deposit and release forms
No full textIntegrative clinical and biomechanical features predict in-hospital trauma mortality
No full textTraumatic injury remains a leading cause of death both in the United States and globally (180,000 and 5 million deaths respectively each year) and there is uncertainty still about which patient clinical parameters impact thrombosis outcomes in these patients and how these clinical and biomechanical alterations together impact mortality subsequent to trauma. The complexity of in-hospital trauma-related mortality is likely the result of an interplay between clinical and biological (pathological) variables and therefore, restrictive single-data-type or step-wise multi-data analyses may not fully capture factors that contribute to mortality and risk. Here, using different state of the art machine learning approaches, our pilot study identifies top 20 integrative features from both clinical (hospital-derived) and biomechanical (laboratory-derived) measurements that help predict patient 3-day in-hospital mortality. We discover that the laboratory measures of biomechanical features are just as, if not more, important in determining the severity of patient outcomes post-injury. Features identified both confirm previously known clinical variables (e.g. PT, PTT, D-dimer) as well as offer new candidates (e.g. clot polymerization rate) that together contribute to improved prediction of mortality than current clinical severity scales (e.g. injury severity score ISS or abbreviated injury scale AIS-HEAD). We believe that these early results hold significant potential in future translation of a novel integrative measure of trauma patient outcome prediction.M.S.Includes bibliographical reference