This research aims to optimize the processing of medical data by developing and implementing an efficient distributed computing platform by leveraging machine learning and edge computing. By doing so, we seek to strik...
详细信息
ISBN:
(纸本)9798350304367;9798350304374
This research aims to optimize the processing of medical data by developing and implementing an efficient distributed computing platform by leveraging machine learning and edge computing. By doing so, we seek to strike a balance between the computational requirements of machine learning models and the need to process medical data locally in many mobile medical imaging scenarios, thus addressing the challenges posed by volume, privacy, and security.
The global state of a distributed system is most commonly stored in databases. Databases are responsible for ensuring that the data is available, durable, and correct at some level of consistency. However, this approa...
详细信息
ISBN:
(纸本)9783031809453;9783031809460
The global state of a distributed system is most commonly stored in databases. Databases are responsible for ensuring that the data is available, durable, and correct at some level of consistency. However, this approach has its limitations. The computing nodes responsible for the application tier need to constantly communicate over the network to be able to interact with the global state and this comes with a performance limitation. Our Traquest model concept allows us to manage the data and even provide full ACID properties on the application tier level instead of the database level. In most cases, the data can be stored right next to the computation even in the same runtime environment. Our prototype measurements imply, that in some cases a Traquest model-based distributed system can provide even magnitudes larger throughput than the fastest in-memory databases today. The Traquest model also introduces a workaround for the famous CAP theorem by introducing temporary availability.
Studying distributed computing through the lens of algebraic topology has been the source of many significant breakthroughs during the last 2 decades, especially in the design of lower bounds or impossibility results....
详细信息
The demand for large-scale computing, such as the application of AI in big data analytics and engineering simulations, has been steadily increasing. Meanwhile, many companies and individuals now own smartphones and PC...
详细信息
ISBN:
(纸本)9783031488023;9783031488030
The demand for large-scale computing, such as the application of AI in big data analytics and engineering simulations, has been steadily increasing. Meanwhile, many companies and individuals now own smartphones and PCs, but a significant portion of these IT assets remains underutilised, especially overnight or utilised as computing resources. This paper solves these issues by proposing to use these resources in a distributed system. This system highlights the opportunities for cost-effective and resource-efficient high-performance computing(HPC) solutions by utilising underutilised resources. The original distributed computing system framework SMPDG will be described in this paper.
distributed computing platforms involve multiple processing systems connected through a network and support the parallel execution of applications. They enable huge computational power and data processing with a quick...
详细信息
ISBN:
(数字)9783031505836
ISBN:
(纸本)9783031505829;9783031505836
distributed computing platforms involve multiple processing systems connected through a network and support the parallel execution of applications. They enable huge computational power and data processing with a quick response time. Examples of use cases requiring distributed computing are stream processing, batch processing, and client-server models. Most of these use cases involve tasks executed in a sequence on different computers to arrive at the results. Numerous distributed computing algorithms have been suggested in the literature, focusing on efficiently utilizing compute nodes to handle tasks within a workflow on on-premises setups. Industries that previously relied on on-premises setups for big data processing are shifting to cloud environments offered by providers such as Azure, Amazon, and Google. This transition is driven by the convenience of Platform-as-a-Service offerings scuh as Batch Services, Hadoop, and Spark. These PaaS services, coupled with auto-provisioning and auto-scaling, reduce costs through a Pay-As-You-Go model. However, a significant challenge with cloud services is configuring them with only a single type of machine for performing all the tasks in the distributed workflow, although each task has diverse compute node requirements. To address this issue in this paper, we propose an Intelligent task scheduling framework that uses a classifier-based dynamic task scheduling approach to determine the best available node for each task. The proposed framework improves the overall performance of the distributed computing workflow by optimizing task allocation and utilization of resources. Although Azure Batch Service is used in this paper to illustrate the proposed framework, our approach can also be implemented on other PaaS distributed computing platforms.
In this paper, we explore a distributed setting, where a user seeks to compute a linearly-separable Boolean function of degree M from N servers, each with a cache size M. Exploiting the fundamental concepts of sensiti...
详细信息
ISBN:
(纸本)9798350393194;9798350393187
In this paper, we explore a distributed setting, where a user seeks to compute a linearly-separable Boolean function of degree M from N servers, each with a cache size M. Exploiting the fundamental concepts of sensitivity and influences of Boolean functions, we devise a novel approach to capture the interplay between dataset placement across servers and server transmissions and to determine the optimal solution for dataset placement that minimizes the communication cost. In particular, we showcase the achievability of the minimum average joint sensitivity, N/2(M-1), as a measure for the communication cost.
Designing high-performance scientific applications has become a time-consuming and complex task that requires developers to master multiple frameworks and toolchains. Although re-configurability and energy efficiency ...
详细信息
ISBN:
(数字)9783031725678
ISBN:
(纸本)9783031725661;9783031725678
Designing high-performance scientific applications has become a time-consuming and complex task that requires developers to master multiple frameworks and toolchains. Although re-configurability and energy efficiency make FPGA a powerful accelerator, efficiently integrating multiple FPGAs into a distributed cluster is a complex and cumbersome task. Such complexity grows considerably when applications require partitioning execution among CPUs, GPUs, and FPGAs. This paper introduces FPGA offloading support to OpenMP cluster (OMPC), an OpenMP-only framework capable of transparently offloading computation across nodes in a cluster, which reduces developer effort and time to solution. In addition, OMPC enables true heterogeneity by allowing the programmer to assign program kernels to the most appropriate architecture (CPUs, GPUs, or FPGA), depending on their workload characteristics. This is achieved by adding only a few lines of standard OpenMP code to the application. The resulting framework was applied to the heterogeneous acceleration of an image recoloring application. Experimental results demonstrate speed-ups gains using different acceleration arrangements with CPU, GPU and FPGA. Measurements using Halstead metrics show that the proposed framework is faster to program. Furthermore, the solution enables transparently offloading OMPC communication tasks to multiple FPGAs, which results in speed-ups of up to 1.41x over the default communication mechanism (Message Passing Interface - MPI) on Task Bench, a synthetic benchmark for task parallelism.
In this paper, we investigate the problem of multi-user linearly decomposable function computation, where N servers help compute functions for K users, and where each such function can be expressed as a linear combina...
详细信息
ISBN:
(纸本)9798350382853;9798350382846
In this paper, we investigate the problem of multi-user linearly decomposable function computation, where N servers help compute functions for K users, and where each such function can be expressed as a linear combination of L basis subfunctions. The process begins with each server computing some of the subfunctions, then broadcasting a linear combination of its computed outputs to a selected group of users, and finally having each user linearly combine its received data to recover its function. As it has become recently known, this problem can be translated into a matrix decomposition problem F = DE, where F is an element of GF(q)(KxL) describes the coefficients that define the users' demands, where E is an element of GF(q)(NxL) describes which subfunction each server computes and how it combines the computed outputs, and where D is an element of GF(q)(KxN) describes which servers each user receives data from and how it combines this data. Our interest here is in reducing the total number of subfunction computations across the servers (cumulative computational cost), as well as the worst-case load which can be a measure of computational delay. Our contribution consists of novel bounds on the two computing costs, where these bounds are linked here to the covering and packing radius of classical codes. One of our findings is that in certain cases, our distributed computing problem - and by extension our matrix decomposition problem - is treated optimally when F is decomposed into a parity check matrix D of a perfect code, and a matrix E which has columns as the coset leaders of this same code.
In this paper, we propose a novel GPU cache system for COMPSs, a task-based distributed computing framework that enables the execution of parallel applications on heterogeneous clusters. GPU COMPSs tasks can exploit t...
详细信息
ISBN:
(纸本)9783031695827;9783031695834
In this paper, we propose a novel GPU cache system for COMPSs, a task-based distributed computing framework that enables the execution of parallel applications on heterogeneous clusters. GPU COMPSs tasks can exploit the computational power of NVIDIA GPUs to process large data blocks. However, the current implementation of COMPSs requires each task to write its output data to disk and the subsequent tasks to read them from disk, which introduces significant overhead. To overcome this limitation, we design and implement a GPU cache system that allows tasks to store and retrieve data from the GPU memory, avoiding unnecessary disk operations and reducing data transfer time. We conducted extensive experiments on several benchmarks and demonstrated that our GPU cache system can achieve significant speedups compared to the baseline COMPSs implementation.
Several algorithms and tools that operate on graphs can significantly benefit from distributed computing. For instance, consider a logistic transportation network represented as a temporal graph. Optimizing transporta...
详细信息
ISBN:
(数字)9783031592058
ISBN:
(纸本)9783031592041;9783031592058
Several algorithms and tools that operate on graphs can significantly benefit from distributed computing. For instance, consider a logistic transportation network represented as a temporal graph. Optimizing transportation routes and times is a well-known NP-hard problem. One typical approach is problem decomposition, which requires optimal partitioning of the network. In such problems, the goals include minimizing the number of cross-partition edges, balancing the sizes of partitions, and controlling the number of partitions to match the capabilities of the computing environment. In this paper, we propose a mathematical formulation of the graph clustering problem for distributed computing environments, along with a simple initial heuristic that can be used to obtain partitions.
暂无评论