2020:

# Project ID Project Description Led By
Availability of large quantities of renewable energy can break the cost curve for large scale computing. For most high performance computing centers, energy is a significant fraction of operating cost, and can reach up to approximately the same cost of com puting clusters over the lifetime of the equipment. Wind power and solar energy are increasingly available in the US and worldwide, but unlike previous sources of renewable energy such as hydroelectric plants, each of these has significantly variable availability and cost (sometimes even negative) throughout the day. To make best use of these sources of energy, data centers will need to be highly automated and preferably sited in remote locations near these sources to reduce transmission costs. This project will apply previous CAC work in data center automation, analytics, and control standards and methods to this problem.
Datacenter-level disaster happens but is rarely studied. Since the consequences of datacenter-level failure could be catastrophic and traditional fault-tolerance techniques will not work, it is critical to understand how to mitigate such disaster without affecting user experiences. The goal of this project is to explore datacenter-level disaster recovery techniques with minimal user impact.
This project aims to develop accelerated solutions of Somewhat Homomorphic Encryption (SHE) based on the Ring Learning with Errors scheme (RLWE). Recent research has identified data movement between host and device as a critical performance bottleneck in RLWE-based SHE. Our work will leverage the recently introduced Unified Memory technology and the Coherent Accelerator Processor Interface (CAPI) to develop an algorithm for the organization and placement of data within a Hybrid Memory System (HMS). The algorithm will reduce the amount of data movement between the host and the device and overlap remote fetches with computation on the accelerator to minimize exposed latency. The decision of when to move data from host to device will be guided by a Machine Learning model. A semi-supervised classifier will be trained offline to learn the relationship between SHE computation and data access characteristics and placement configurations. The optimal configuration will then be selected at runtime via autotuning methods. In prior work, we have deployed ML-driven heuristics to optimize accelerated applications where it has yielded integer factor performance improvements.
The advent of ‘smart’ infrastructure systems that integrate digital communications and controls, with human operators as beneficiaries have created more new vulnerabilities than would exist if the sub-systems were isolated from one another. Sophisticated cyberattacks can exploit these vulnerabilities to disrupt or even completely disable the operations of our critical infrastructures and their services. The recent embrace of Internet of Things (IoT), autonomous driving, and cloud computing will further exacerbate the cybersecurity problem. Anomaly based Intrusion Detection Systems (ABA-IDS) are the go-to approach to detect and secure this smart infrastructure. ABA-IDS heavily depend on accurate modeling of the normal behavior of the systems to detect attacks. This modeling of the normal behavior is performed using machine learning techniques. To address IoT security problem not only do we need to train models that are highly accurate in anomaly detection but also train models and use machine learning techniques that are highly scalable in order to detect abnormal behavior in large cluster of IoT devices. The main goal of this research project is to explore the use of GPUs and other parallel processing techniques to improve the performance (computational) of machine learning models to aid in secure IoT infrastructure.
There is an increasing dependency of individual users, as well as businesses, on secure computers and computer networks. Furthermore, there is also an increase in the threats to them by malicious agents. Thus, there is an urgent need for accurate, adaptive and automatic detection of cyber-attacks and intrusions has emerged. The ever-increasing need for the adaptive, autonomic detection of and protection against attacks/intrusions is evident. In this project we propose develop a modular and adaptive cyber immunity mechanism to overcome security deficiencies of current computing systems.
Disk arrays spread data across several disks and access them in parallel to increase data transfer rates and I/O rates. Disk arrays are, however, highly vulnerable to disk failures. One of the challenges facing RAID storage technology is the growing time needed to rebuild failed disks, which increases the risk of data loss and threatens the long-running data storage technology's viability. When a disk fails, one RAID controller and a handful of disks do all the recovery work while the other disks and RAID controllers are not involved in the recovery process. The time it takes to rebuild an 8 TB or larger disk drive can reach several days, depending on how busy the storage system and RAID group are. In this project, we will explore declustered RAID, machine learning enabled proactive disk data protection, and vectorized RAIDZ technologies on both HDD and SSD drives to develop always-on HPC storage systems.
# Project ID Project Description Led By
Monitoring data centers is challenging due to their size, complexity, and dynamic nature. This project proposes visual approaches for situational awareness and health monitoring of HPC systems. The visualization requirements are expanded on the following dimensions: HPC spatial layout, temporal domain (historical vs. real-time tracking), jobs scheduling data, and system health services such as temperature, fan speed, and power consumption. We therefore focus the following design goals: 1) Provides spatial and temporal overview across hosts and racks, 2) Allows system administrators to filter by time series features such as sudden changes in temperatures for system debugging,3) Inspects the correlation of system health services and job scheduling information in a single view and 4) Characterize the HPC systems using unsupervised learning and provide predictive analysis.
With the collaboration between Tactical Computing Labs(TCL), Texas Tech Univ. (TTU), and Boston Univ.(BU),the goals of the Extended Base Global Address Space (xBGAS) project are three-fold:1)toprovide128-bitextended addressing capabilities based on RISC-V architecture with ABI compatibility(e.g.RV64 apps will execute without an issue),2)extended addressing must be flexible to support multiple target application spaces/system architectures, e.g. traditional data centers, clouds, HPC, etc.,and3)extended addressing must not rely upon any one virtual memory mechanism. The overall xBGAS project includes four major tasks: design and development of ISA extension, design and development of compiler and tool chain to support extended instructions and addressing modes, design and development of runtime libraries, and porting benchmarks and applications
Data recovery and salvage of all county booking, court proceedings, probation and mental health records. Data is being recovered from a 1989 unstable system. We will compile data and return 30-years of recovered data.
The primary goal of this project is to design and develop the capability of an intelligent assistant that can be used primarily in the health care field. We are going to utilize different technologies, such as natural language processing (NLP), scenario representation, AI-based chat capability, and detection of cognitive state. The intelligent assistant can be trained to perform in different roles including a personal assistant of a patient, a nurse, or it can even play the role of a patient in a training environment for doctors and nurses.
Delirium affects 60% of hospitalized dementia patients, impacting long-term survival and quality of life. SeVA (Senior’s Virtual Assistant) aims to address the potential gaps of the current hospital practices in early delirium detection, management, and prevention through continuous monitoring of clinical, mental, and emotional factors. Thus, SeVA provides timely non-pharmacological intervention for patients with Alzheimer’s disease and related dementias.
Cyberspace includes a wide range of physical networks, storage and computing devices, applications, and users with different roles and requirements. Securing and protecting such complex and dynamic cyberspace resources and services are grand challenges. MLABA(Multi-Layer Anomaly Behavior Analysis) aims at developing a multi-layer anomaly behavior analysis of all components associated with each cyberspace layer and how they interact with each other in order to achieve superior capabilities in characterizing their normal operations and proactively detect any anomalous behavior that might be triggered by malicious attacks. MLABA uses unsupervised deep learning to collect the feature sets of Physical Persona Footprint (PPF), Logical Persona Footprint (LPF), and User Persona Footprint (UPF), to classify normal and abnormal behaviors.
With the advances in the Internet technologies and services, the social media has been gaining excessive popularity and these technologies provide anonymity which harbors hacker discussion forums, underground markets, dark web, etc. We propose to use unsupervised author identification techniques for Twitter to tackle social media forensics cases when suspects of authorship are either missing or not reliable. We will develop the tools to collect potential malicious twitter user data and extract unique signatures that can be used to model the characteristics of Twitter users and design unsupervised learning-based method to identify suspects in Twitter.
With the rapid deployment of IoT devices, Personal Area Networks (PAN)such as Bluetooth networks have become the wireless network choice for small range communications. It is important that Bluetooth networks are secure against cyberattacks like Denial of Service (DoS), Man-in-the Middle attack, battery draining attacks, etc. In this project we will develop an anomaly-based intrusion detection system for Bluetooth networks (BIDS). The BIDS will use an n-gram approach to characterize the normal behavior of the Bluetooth protocol. This project will help in detecting malicious attacks like power utilization attack, DoS, man-in-the-middle attack, etc. on the Bluetooth network.
The advent of ‘smart’ infrastructure systems has created new. Sophisticated cyberattacks can exploit these vulnerabilities to disrupt or even completely disable the operations of our critical infrastructures and their services. There are many possible testbeds (physical and virtual) and simulations for critical infrastructures and cyber systems. To understand the interdependency among these testbeds and their implications on cybersecurity issues, it is important to be able to compose several testbeds into one federated testbed that includes smart devices and sensors, IoT devices, cloud systems, smart grids, smart buildings, etc. (ultimately what is known as smart cities or smart governments). The main goal of this project is to explore innovative techniques to allow seamless composition of a federated testbed that consists of several heterogeneous testbeds include virtual cybersecurity testbeds, IoT testbeds, and cyber-physical testbeds.

2018:

# Project ID Project Description Led By
The focus of this work is on creation of testbeds that can allow direct operation and comparison of a variety of cloud software. This work is being conducted in cooperation with the Cloud Plugfest developer testing series, US National Institute of Standards and Technology (NIST) cloud computing efforts, and related cloud software testing programs. This testing is used to discover, itemize and understand several specific aspects of the cloud computing landscape to identify approaches of business value to CAC members. This project can be expanded to include other developing standards testing as needed.
The focus of this work is on creation of testbeds that can allow direct operation and comparison of a variety of cloud software. This work is being conducted in cooperation with the Cloud Plugfest developer testing series, US National Institute of Standards and Technology (NIST) cloud computing efforts, and related cloud software testing programs. This testing is used to discover, itemize and understand several specific aspects of the cloud computing landscape to identify approaches of business value to CAC members. This project can be expanded to include other developing standards testing as needed.

2017:

# Project ID Project Description Led By
The focus of this work is on creation of testbeds that can allow direct operation and comparison of a variety of cloud software.This work is being conducted in cooperation with the Cloud Plugfest developer testing series, US National Institute of Standards and Technology (NIST) cloud computing efforts, and related cloud software testing programs. This testing is used to discover, itemize and understand several specific aspects of the cloud computing landscape to identify approaches of business value to CAC members. This project can be expanded to include other developing standards testing as needed.
The focus of this work is on creation of testbeds that can allow direct operation and comparison of a variety of cloud software. This work is being conducted in cooperation with the Cloud Plugfest developer testing series, US National Institute of Standards and Technology (NIST) cloud computing efforts, and related cloud software testing programs. This testing is used to discover, itemize and understand several specific aspects of the cloud computing landscape to identify approaches of business value to CAC members. This project can be expanded to include other developing standards testing as needed.
This project concentrates on significantly enhancing data durability in data centers without using remote-site data replication (geo-replication) like in existing solutions. It also considers other important characteristics of the storage system including load balance and scalability holistically. One of the major research challenges is that some of the existing techniques of improving the correlated-failure durability sacrifices load balance, scalability, and numerous other features of the existing storage systems. This project will investigate these issues and create innovative solutions.
Big data often involves disparate sources of data from unrelated files that are available over a period of many years. The presumably unrelated data, however, is difficult to organize and use due to its size and dissimilar attributes having different factor nomenclature and/or file formats. The main purpose of this project is to enable technologies to organize, visualize, and analyze large repositories of potentially unrelated files across several formats. The products from this work will be integrated into precision risk Big Data analytics through the center.
The proposed project refines risk algorithms developed for handling disparate sources of data. This effort extends project 2016-TTU-5
A common trend in Intrusion Detection Systems (IDSs) is to consider data structures based on graphs to analyze network traffic and attack patterns. Timely detecting a threat is fundamental to reduce the risk to which the system is exposed, but no current studies aim at providing useful information to size Cloud or HPC infrastructures to meet certain service level objectives. This project targets to prototype a completely distributed property-graph based benchmark for next generation IDSs.
The current cloud security techniques are mainly labor intensive, use signature based attack detection tools, and not flexible enough to handle the current cyberspace complexity, dynamism, and epidemic-style propagation of attacks. Furthermore, while the organization boundaries are gradually disappearing, it became infeasible to create a defendable perimeter. In addition, the insider attacks are still a major issue since they have access to the cloud services and even infrastructures. In this project, scalable and resilient cloud services will be prototyped and tested using a public cloud service such as Amazon AWS.
The main purpose of this project is to prototype an HCI environment for Chronic Heart Failure (CHF) that guides clinical experiments, and improves care delivery and patient management.
The main goal of this project is to design and develop a IoT Security Framework to enable developers to predict and mitigate security issues in a systematically. The project evaluates a security framework to deploy highly secure IoT applications for Smart Infrastructures. Our approach is based on an Anomaly Behavior Analysis methodology to detect any type of attack (known or unknown).

2016:

# Project ID Project Description Led By
Benchmarking of a system is a process of assessing its performance and other characteristics, to be able to compare them with other systems. There is a strong desire for benchmarking and evaluating Cloud computing systems. Cloud computing provides service-oriented access to computing, storage and networking resource. On one hand, considering the number of Cloud computing providers and the different services each provider offers, Cloud users need benchmark information to select the best service and provider for their needs. On the other hand, Cloud providers and Cloud architects need benchmarking results to create optimized architectures. The problem is that Cloud is a very complex and dynamic environment. Therefore there are important differences between benchmarking in dynamic Cloud environment with traditional benchmarking methods in static systems. We have performed an initial research about benchmarking in Cloud and understand various challenges about creating and deploying Cloud benchmarking; and we found that the most important problem for Cloud users are data isolation, security and unreliable performance. Inference and performance isolation are other major challenges for developers and architects of Cloud systems.
Emerging large-scale applications on Cloud computing platform, such as information retrieval, data mining, online business, and social network, are data- rather than computation-intensive. Storage system is one of the most critical components for Cloud computing. The traditional hard disk drives (HDD) are current dominant storage devices in Clouds, but are notorious for long access latency and failure prone. The emerging storage class memory (SCM) such as Solid State Drives provides a new promising storage solution of high bandwidth, low latency, and mechanical component free, but with inherent limitations of small capacity, short lifetime, and high cost. The objective of this project is to build an innovative unified storage architecture (Unistore) with the co-existence and efficient integration of heterogeneous HDD and SCM devices for Cloud storage systems.
Financial Intelligence for Banks is a new project of the TTU Cloud Computing efforts. The focus of this work is on creation of software and testing methodologies that can allow banks the ability to predict short term market analysis for customer portfolios through stress testing, loan product pricing, and target marketing for current and potential customers. This work is being conducted in cooperation with CAC partner site Happy State Bank in Amarillo, Texas. Dr. Perez has been project leader of this working group for the CAC since July 2015. Center for Advanced Analytics and Business Intelligence (CAABI) is a major participating institution at TTU based its existing research portfolio.
This project combines individual and public health exposome data to develop and prototype refined disease risk analytics using multi-level models. These models will then be validated using obesity and CVD exemplar datasets, as well as Patient Health Information (PHI).
The current cloud security techniques are mainly labor intensive, use signature based attack detection tools, and not flexible enough to handle the current cyberspace complexity, dynamism, and epidemic-style propagation of attacks. Furthermore, while the organization boundaries are gradually disappearing, it became infeasible to create a defendable perimeter. In addition, the insider attacks are still a major issue since they have access to the cloud services and even infrastructures. In this project, scalable and resilient cloud services will be prototyped and tested using a public cloud service such as Amazon AWS."

2015:

# Project ID Project Description Led By
Benchmarking of a system is a process of assessing its performance and other characteristics, to be able to compare them with other systems. There is a strong desire for benchmarking and evaluating Cloud computing systems. Cloud computing provides service-oriented access to computing, storage and networking resource. On one hand, considering the number of Cloud computing providers and the different services each provider offers, Cloud users need benchmark information to select the best service and provider for their needs. On the other hand, Cloud providers and Cloud architects need benchmarking results to create optimized architectures. The problem is that Cloud is a very complex and dynamic environment. Therefore there are important differences between benchmarking in dynamic Cloud environment with traditional benchmarking methods in static systems. We have performed an initial research about benchmarking in Cloud and understand various challenges about creating and deploying Cloud benchmarking; and we found that the most important problem for Cloud users are data isolation, security and unreliable performance. Inference and performance isolation are other major challenges for developers and architects of Cloud systems.
Emerging large-scale applications on Cloud computing platform, such as information retrieval, data mining, online business, and social network, are data- rather than computation-intensive. Storage system is one of the most critical components for Cloud computing. The traditional hard disk drives (HDD) are current dominant storage devices in Clouds, but are notorious for long access latency and failure prone. The emerging storage class memory (SCM) such as Solid State Drives provides a new promising storage solution of high bandwidth, low latency, and mechanical component free, but with inherent limitations of small capacity, short lifetime, and high cost. The objective of this project is to build an innovative unified storage architecture (Unistore) with the co-existence and efficient integration of heterogeneous HDD and SCM devices for Cloud storage systems.
The main purpose of this project is to prototype readmission risk and insurance contract analytics This project involves deploying a sophisticated data warehouse for hosting, managing and processing large volumes of healthcare data from disparate sources. The secondary objective is to prototype-advanced analytics that helps in risk stratification and assessing risk category transition in population health modeling. The tertiary objective is to model insurance contracts and their business value.

2014:

# Project ID Project Description Led By
Benchmarking of a system is a process of assessing its performance and other characteristics, to be able to compare them with other systems. There is a strong desire for benchmarking and evaluating Cloud computing systems. Cloud computing provides service-oriented access to computing, storage and networking resource. On one hand, considering the number of Cloud computing providers and the different services each provider offers, Cloud users need benchmark information to select the best service and provider for their needs. On the other hand, Cloud providers and Cloud architects need benchmarking results to create optimized architectures. The problem is that Cloud is a very complex and dynamic environment. Therefore there are important differences between benchmarking in dynamic Cloud environment with traditional benchmarking methods in static systems. We have performed an initial research about benchmarking in Cloud and understand various challenges about creating and deploying Cloud benchmarking; and we found that the most important problem for Cloud users are data isolation, security and unreliable performance. Inference and performance isolation are other major challenges for developers and architects of Cloud systems.
Emerging large-scale applications on Cloud computing platform, such as information retrieval, data mining, online business, and social network, are data- rather than computation-intensive. Storage system is one of the most critical components for Cloud computing. The traditional hard disk drives (HDD) are current dominant storage devices in Clouds, but are notorious for long access latency and failure prone. The emerging storage class memory (SCM) such as Solid State Drives provides a new promising storage solution of high bandwidth, low latency, and mechanical component free, but with inherent limitations of small capacity, short lifetime, and high cost. The objective of this project is to build an innovative unified storage architecture (Unistore) with the co-existence and efficient integration of heterogeneous HDD and SCM devices for Cloud storage systems.
The main purpose of this project is to prototype readmission risk and insurance contract analytics This project involves deploying a sophisticated data warehouse for hosting, managing and processing large volumes of healthcare data from disparate sources. The secondary objective is to prototype-advanced analytics that helps in risk stratification and assessing risk category transition in population health modeling. The tertiary objective is to model insurance contracts and their business value.