位置感知的Web服务性能预测研究
发布时间：2018-10-29 16:11
【摘要】：Web服务的性能是用户在选择Web服务时常会考虑的一个重要因素。用户对未访问过的服务性能并不清楚。因此在选择服务时，常常需要对它们的性能进行预测，来帮助用户选择到满意的服务。如何提高Web服务性能的预测精度是近年来的一个热点研究问题。最近，一些工作考虑了用户或者Web服务的位置对服务性能的影响，从理论上提出了位置感知的Web服务性能预测方法。然而，很少有工作在实践中针对Web服务性能与用户、服务位置的关系进行验证。 本文在考虑地理位置和网络位置的前提下，提出了一套对Web服务性能与用户、服务位置的关系进行验证的方法，，并使用了大规模的真实Web服务的性能数据集进行了实验。基于地理位置的方法通过映射IP地址到经纬度，计算出用户（服务）之间的地理距离；然后分别验证了地理位置对服务性能的影响和用户（服务）之间的地理距离对它们之间相似度的影响。基于网络位置的方法定义在相同国家或AS的用户（服务）为邻居，通过将IP地址映射到相应的国家或AS；然后分别验证了网络位置对服务性能的影响和用户（服务）之间的网络距离对它们之间相似度的影响。 以上两种方法的实验结果均表明：（1）同一个用户调用不同的服务时，位置近的服务性能均值要优于远处的服务性能均值；（2）同一个服务被不同的用户调用时，距离近的用户调用的服务性能要更好；（3）位置近的用户（服务）相似度要高于远处的用户（服务）相似度；（4）如果两个用户（服务）具有相似的性能，那么这两个用户（服务）有很高的概率在位置上也临近；反过来亦如此。通过对比上述两种方法发现，网络位置比地理位置与服务性能的关系更密切。在实验结果的基础上，文中还提出了几种利用地理位置和网络位置的Web服务性能预测方法，与已有的一些经典的Web服务性能预测方法相比，在一定程度上切实提高了服务性能的预测精度。
[Abstract]:The performance of Web services is an important factor that users often consider in choosing Web services. Users are not aware of unaccessed service performance. Therefore, when selecting services, we often need to predict their performance to help users select satisfactory services. How to improve the prediction accuracy of Web service performance is a hot issue in recent years. Recently, some work has considered the influence of user or Web service location on service performance, and put forward a location-aware Web service performance prediction method in theory. However, little work has been done in practice to verify the relationship between Web service performance and user, service location. On the premise of considering geographical location and network location, this paper proposes a method to verify the relationship between Web service performance and user, service location, and uses a large scale real Web service performance data set to carry out experiments. The geographical distance between users (services) is calculated by mapping IP address to latitude and longitude. Then the effects of geographical location on service performance and geographic distance between users (services) on their similarity were verified. A network-based approach is defined as a neighbor to a user (service) in the same country or AS by mapping the IP address to the corresponding country or AS; Then the influence of network location on service performance and the influence of network distance between users (services) on their similarity are verified respectively. The experimental results of the above two methods show that: (1) when the same user invokes different services, the average value of service performance near the location is better than the average service performance in the distance; (2) when the same service is called by different users, the service performance of the close user calls is better, (3) the similarity of the user (service) near the location is higher than that of the distant user (service). (4) if two users (services) have similar performance, then the two users (services) have a high probability of approaching in the location and vice versa. By comparing the above two methods, it is found that network location is more closely related to service performance than geographical location. Based on the experimental results, several Web service performance prediction methods based on geographical location and network location are proposed, which are compared with some classical Web service performance prediction methods. The prediction accuracy of service performance is improved to a certain extent.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09
本文编号：2298176
[Abstract]:The performance of Web services is an important factor that users often consider in choosing Web services. Users are not aware of unaccessed service performance. Therefore, when selecting services, we often need to predict their performance to help users select satisfactory services. How to improve the prediction accuracy of Web service performance is a hot issue in recent years. Recently, some work has considered the influence of user or Web service location on service performance, and put forward a location-aware Web service performance prediction method in theory. However, little work has been done in practice to verify the relationship between Web service performance and user, service location. On the premise of considering geographical location and network location, this paper proposes a method to verify the relationship between Web service performance and user, service location, and uses a large scale real Web service performance data set to carry out experiments. The geographical distance between users (services) is calculated by mapping IP address to latitude and longitude. Then the effects of geographical location on service performance and geographic distance between users (services) on their similarity were verified. A network-based approach is defined as a neighbor to a user (service) in the same country or AS by mapping the IP address to the corresponding country or AS; Then the influence of network location on service performance and the influence of network distance between users (services) on their similarity are verified respectively. The experimental results of the above two methods show that: (1) when the same user invokes different services, the average value of service performance near the location is better than the average service performance in the distance; (2) when the same service is called by different users, the service performance of the close user calls is better, (3) the similarity of the user (service) near the location is higher than that of the distant user (service). (4) if two users (services) have similar performance, then the two users (services) have a high probability of approaching in the location and vice versa. By comparing the above two methods, it is found that network location is more closely related to service performance than geographical location. Based on the experimental results, several Web service performance prediction methods based on geographical location and network location are proposed, which are compared with some classical Web service performance prediction methods. The prediction accuracy of service performance is improved to a certain extent.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09
【参考文献】
相关期刊论文 前5条
1 刘国奇;朱志良;刘莹;王琴;;一种基于服务使用信息的Web服务QoS度量方法[J];东北大学学报(自然科学版);2009年10期
2 冯在文;何克清;李兵;龚平;何扬帆;刘玮;;一种基于情境推理的语义Web服务发现方法[J];计算机学报;2008年08期
3 邓水光;尹建伟;李莹;吴健;吴朝晖;;基于二分图匹配的语义Web服务发现方法[J];计算机学报;2008年08期
4 韩忠民;;知经纬度计算两点精确距离[J];科技传播;2011年11期
5 唐明董;姜叶春;刘建勋;;用户位置感知的Web服务QoS预测方法[J];小型微型计算机系统;2012年12期
本文编号：2298176
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2298176.html
