A5026624531- Cloud Computing and Resource Management
- Distributed and Parallel Computing Systems
- Parallel Computing and Optimization Techniques
- Scientific Computing and Data Management
- Advanced Data Storage Technologies
- IoT and Edge/Fog Computing
- Digital Transformation in Industry
- Distributed systems and fault tolerance
- Radiation Effects in Electronics
- Quantum Information and Cryptography
- Augmented Reality Applications
- Optimization and Packing Problems
- Advancements in Semiconductor Devices and Circuit Design
- Caching and Content Delivery
- Quantum Computing Algorithms and Architecture
- Scheduling and Optimization Algorithms
- Blockchain Technology Applications and Security
- Stochastic Gradient Optimization Techniques
Hewlett-Packard (United States)
2023-2024
Hewlett Packard Enterprise (United States)
2024
Binghamton University
2016-2019
Cloud platforms typically require users to provide resource requirements for applications so that managers can schedule containers with adequate allocations. However, the container resources often depend on numerous factors such as application input parameters, optimization flags, files, and attributes are specified each run. So, it is complex estimate a given accurately, leading over-estimation negatively affects overall utilization. We have designed Resource Utilization Based Autoscaling...
Sustainability has become a critical problem confronting the community. Ecological issues such as climate change and CO2 emissions, economical frictions energy supplies, socio-political wars threaten growth equity. How can technology help?
Traditional high-performance computing and modern artificial intelligence are converging with workflows as a common paradigm. We predict nine principles of heterogeneity serverless for this convergence, from high-level programming to low-level hardware.
Both commercial clouds and academic campus clusters suffer from low resource utilization long wait times as the estimates for jobs, provided by users, is often inaccurate. Incorrect estimation poses challenges in overall cluster cloud management. Under allocation can cause significant slowdown or termination of applications. Over-allocation resources applications causes increased pending tasks queue, reduced throughput, underutilization cluster. For end users that pay allocations, incorrect...
Cloud infrastructures increasingly include a heterogeneous mix of components in terms performance, power, and energy usage. As the size cloud grows, power consumption becomes significant constraint. We use Apache Mesos Aurora, which provide massive scalability to Web-scale applications, demonstrate how policy driven approach involving bin-packing workloads according their profiles, instead default allocation by can effectively reduce peak-power usage as well node utilization, when are...
The growing convergence of high-performance, data analytics, and machine-learning applications is increasingly pushing computing systems toward heterogeneous processors specialized hardware accelerators. Hardware heterogeneity, in turn, leads to finer-grained workflows. State-of-the-art server-less resource managers do not currently provide efficient scheduling such fine-grained tasks on with CPUs accelerators (e.g., GPUs FPGAs). Working presents an opportunity for more energy use via new...
Application energy efficiency can be improved by executing each application component on the compute element that consumes least while also satisfying time constraints. In principle, function as a service (FaaS) paradigm should simplify such optimizations abstracting away location, but existing FaaS systems do not provide for user transparency over consumption or task placement. Here we present GreenFaaS, novel open source framework bridges this gap between energy-efficient applications and...
Function-as-a-service (FaaS) is a promising execution environment for high-performance computing (HPC) and machine learning (ML) applications as it offers developers simple way to write deploy programs. Nowadays, GPUs other accelerators are indispensable HPC ML workloads. These expensive acquire operate; consequently, multiplexing them can increase their financial profitability. However, we have observed that state-of-the-art FaaS frameworks usually treat accelerator single device run...
As resource estimation for jobs is difficult, users often overestimate their requirements. Both commercial clouds and academic campus clusters suffer from low utilization long wait times as the estimates jobs, provided by users, inaccurate. We present an approach to statistically estimate actual requirement of a job in Little cluster before run Big cluster. The initial on little gives us view how much resources requires. This allows accurately allocate pending queue thereby improve...
Summary Since the dawn of quantum computing (QC), theoretical developments like Shor's algorithm proved conceptual superiority QC over traditional computing. However, such supremacy claims are difficult to achieve in practice because technical challenges realizing noiseless qubits. In near future, applications will need rely on noisy devices that offload part their work classical devices. One way this is by using parameterized circuits optimization or even machine learning tasks. The energy...
With the slowing of Moore's law and decline Dennard scaling, computing systems increasingly rely on specialized hardware accelerators in addition to general-purpose compute units. Increased heterogeneity necessitates disaggregating applications into workflows fine-grained tasks that run a diverse set CPUs accelerators. Current accelerator delivery models cannot support such efficiently, as (1) overhead managing erases performance benefits for tasks; (2) exclusive use per task leads...
Academic cloud infrastructures require users to specify an estimate of their resource requirements. The usage for applications often depends on the input file sizes, parameters, optimization flags, and attributes, specified each run. Incorrect estimation can result in low utilization entire infrastructure long wait times jobs queue. We have designed a Resource Utilization based Migration (RUMIG) system address problem. present overall architecture two-stage elastic cluster design, Apache...