With the development of cloud service technologies, the amount of services grows rapidly, leading to building high-quality services an urgent and crucial research problem. Service users should evaluate QoS to select t...
详细信息
With the development of cloud service technologies, the amount of services grows rapidly, leading to building high-quality services an urgent and crucial research problem. Service users should evaluate QoS to select the optimal cloud services from a series of functionally equivalent service candidates, because QoS performance of services is varying over time. The reason is that QoS is related to the service overload and network environments. This phenomenon makes QoS prediction for users located in different places even harder. Furthermore, since service invocations are charged by service providers, it is impractical to let users invoke required cloud services to evaluate quality with respect to time and resources. To solve this problem, this paper proposes a cloud service QoS prediction method, called TPP (Time-aware and Parallel Prediction), to provide time-aware and parallel QoS value prediction for various service users. TPP is able to predict without additional invocation of cloud services, since it uses past cloud service usage experience from different service users. We propose and implement tensor decomposition algorithm on the Spark system. The results of extensive experimental show the accuracy and efficiency of TPP.
Massive graphs, such as online social networks and communication networks, have become common today. To efficiently analyze such large graphs, many distributed graph computing systems have been developed. These system...
详细信息
ISBN:
(纸本)9781450334693
Massive graphs, such as online social networks and communication networks, have become common today. To efficiently analyze such large graphs, many distributed graph computing systems have been developed. These systems employ the "think like a vertex" programming paradigm, where a program proceeds in iterations and at each iteration, vertices exchange messages with each other. However, using Pregel's simple message passing mechanism, some vertices may send/receive significantly more messages than others due to either the high degree of these vertices or the logic of the algorithm used. This forms the communication bottleneck and leads to unbalanced workload among machines in the cluster. In this paper, we propose two effective message reduction techniques: (1)vertex mirroring with message combining, and (2)an additional request respond API. These techniques not only reduce the total number of messages exchanged through the network, but also hound the number of messages sent/received by any single vertex. We theoretically analyze the effectiveness of our techniques, and implement them on top of our open -source Pregel implementation called Pregel+. Our experiments on various large real graphs demonstrate that our message reduction techniques significantly
暂无评论