Nowadays, Peer-to-Peer technology has been widely used in live streaming applications and many related systems are proposed. However, their single overlay design and unbalanced scheduling methods lead to some ineffici...
详细信息
ISBN:
(纸本)9781595937575
Nowadays, Peer-to-Peer technology has been widely used in live streaming applications and many related systems are proposed. However, their single overlay design and unbalanced scheduling methods lead to some inefficiency including high control overhead and bad playback experience. This paper mainly discusses how to address these certain problems. We introduce a hybrid architecture to solve the locality problem and reduce the control overhead. We also propose a scheduling method to achieve load balance. All solutions are implemented in Anysee2.
Cloud provides users with a new model of utilizing the computing infrastructure with the ability to perform parallel and distributed computations using elastic virtual cluster. However, the multi-level and complex fea...
详细信息
Silicosis remains one of the most harmful occupational respiratory diseases. It threatens the workers exposed to dust environment. Chest radiograph is the main available image source for silicosis diagnosis according ...
详细信息
Non-uniform memory access (NUMA) is one of the main architectures of today’s high-performance server. The key feature of NUMA is the non-uniformity of access latency. Access from a processor to attached memory is fas...
详细信息
Traditionally, complex engineering applications （CEAs）, which consist of numerous components （software） and require a large amount of computing resources, usu- ally run in dedicated clusters or high performance co...
详细信息
Traditionally, complex engineering applications （CEAs）, which consist of numerous components （software） and require a large amount of computing resources, usu- ally run in dedicated clusters or high performance computing （HPC） centers. Nowadays, Cloud computingsystem with the ability of providing massive computing resources and cus- tomizable execution environment is becoming an attractive option for CEAs. As a new type on Cloud applications, CEA also brings the challenges of dealing with Cloud resources. In this paper, we provide a comprehensive survey of Cloud resource management research for CEAs. The survey puts forward two important questions： 1） what are the main chal- lenges for CEAs to run in Clouds？ and 2） what are the prior research topics addressing these challenges？ We summarize and highlight the main challenges and prior research topics. Our work can be probably helpful to those scientists and en- gineers who are interested in running CEAs in Cloud envi- ronment.
Multi-instance learning (MIL) is a celebrated learning framework where each example is represented as a bag of instances. An example is negative if it has no positive instances, and vice versa if at least one positive...
详细信息
Knowledge representation learning is a key step required for link prediction tasks with knowledge graphs (KGs). During the learning process, the semantics of each entity are embedded by a vector or a point in a featur...
详细信息
Efficient metadata management is a critical aspect of overall system performance in large distributed storage systems. Directory subtree partitioning and traditional hashing are two common techniques used for managing...
详细信息
The volume of RDF data continues to grow over the past decade and many known RDF datasets have billions of triples. A grant challenge of managing this huge RDF data is how to access this big RDF data efficiently. A po...
详细信息
Multi-hop reasoning has been widely studied for its important application values in the domain of intelligent search and question answering. Real-world applications are often dominated by natural language input, and i...
详细信息
暂无评论