Group-user access patterns and tile prefetching based on a time-sequence distribution in Cloud-based GIS
Group-user access patterns and tile prefetching based on a time-sequence distribution in Cloud-based GIS
复制标题
云GIS中基于时间序列分布的组用户访问模式和图块预取
DOI：
10.1016/j.compenvurbsys.2017.12.002
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
董广胜
中科院分区：
文献类型：
--
作者：
李锐;樊珈珮;吴华意;蒋捷;董广胜
组用户对空间数据的密集访问在时间上是聚合的，而访问的内容在空间上是相关的，在现有的基于云的GIS（CGIS）中产生性能波动和服务瓶颈。目前的解决方案是不可扩展的，因为它们没有考虑到定量密集的时空访问模式，可以使CGIS更准确，更快地适应聚合和突发组用户访问。在我们的研究中，我们解决了这些性能波动和服务瓶颈问题，一种新的高斯混合模型（GMM）表示短期聚合的瓦片访问。本文提出了一种新的瓦片预取和缓存预取控制策略，量化瓦片访问的时空相关性，以优化CGIS服务性能; GMM算法根据群组用户访问的多峰特性和时间序列分布的均匀变化，量化瓦片访问流行度的变化，预测热点瓦片的访问概率。使用累积访问概率来量化瓦片访问的空间相关性和局部性。在我们的方法中，在一个有效的缓存预取控制策略的基础上，资源利用率的热点瓦片和相邻的瓦片在一个给定的空间区域进行预取。实验结果表明，基于GMM的瓦片访问概率预测方法（TAP-GMM）能够通过稳定的瓦片预取服务准确预测热点瓦片及其访问流行度分布，具有较好的可扩展性。首先，我们提出了GMM短期变化的瓦片访问流行度，固定的空间相关性，和时间局部性的访问瓦片在组用户密集的访问行为。其次，我们对访问行为的研究使得预取策略能够适应不同的强度和突发模式，在组用户访问优化CGIS的性能。预取策略是稳定的，减少了服务资源消耗，因为它符合定量测量的访问模式。
Group-user intensive access to spatial data is temporally aggregated, while accessed content is spatially correlated, creating performance fluctuations and service bottlenecks in existing cloud-based GISs (CGISs). Current solutions are not scalable as they fail to consider the quantitatively intensive spatiotemporal access patterns that could make a CGIS adapt more accurately and quickly to aggregative and bursting group-user accesses. In our research, we tackled these performance fluctuations and service bottleneck problems with a novel Gaussian mixture model (GMM) representing the short-term aggregation of tile accesses. We constructed a new tile prefetching and cache prefetching control strategy that represents quantitatively the spatiotemporal correlations of accessed tiles to optimize CGIS service performance.The GMM solution that we proposed quantifies variation in tile access popularity to predict access probabilities for hotspot tiles based on the observed multipeak characteristics and uniform variations in the time-sequence distribution found in group-user accesses. Cumulative access probability was used to quantify the spatial correlation and locality of tile accesses. In our approach, a hotspot tile and the neighboring tiles in a given spatial area are prefetched in an effective cache prefetching control strategy based on resource utilization. Experiments have demonstrated that the proposed GMM for tile access probability (TAP-GMM) accurately predicts hotspot tiles and their access popularity distributions by prefetching tiles through a stable service with enhanced scalability.The contributions of this study are twofold. First, we present the GMM for short-term variations of tile access popularity, fixed spatial correlation, and temporal locality of accessed tiles in group-user intensive access behavior. Second, our study on access behavior makes the resulting prefetching strategy adaptable to varying intensities and bursting patterns in group-user access for optimizing CGIS performance. The prefetching strategy is stable and reduces the service resource consumption as it conforms to quantitatively measured access patterns.
登录
查看更多内容
DOI：
10.1145/954339.954341
发表时间：
2003-12
期刊：
ACM Comput. Surv.
影响因子：
--
作者：
Stefan Podlipnig;L. Böszörményi
通讯作者：
Stefan Podlipnig;L. Böszörményi
DOI：
10.5120/9342-3659
发表时间：
2012-11
期刊：
International Journal of Computer Applications
影响因子：
--
作者：
K. Noipithuk;C. Kitkobchai;C. Chantrapornchai
通讯作者：
K. Noipithuk;C. Kitkobchai;C. Chantrapornchai
DOI：
10.1080/13658816.2012.721555
发表时间：
2013-03
影响因子：
5.7
作者：
Ricardo García Martín;Juan Pablo de Castro Fernández;E. Verdú;M. Verdú;Luisa María Regueras Santos
通讯作者：
Ricardo García Martín;Juan Pablo de Castro Fernández;E. Verdú;M. Verdú;Luisa María Regueras Santos
DOI：
10.1145/2396761.2398577
发表时间：
2012-10
期刊：
Proceedings of the 21st ACM international conference on Information and knowledge management
影响因子：
--
作者：
Daniar Achakeev;B. Seeger;P. Widmayer
通讯作者：
Daniar Achakeev;B. Seeger;P. Widmayer
DOI：
10.1016/j.compenvurbsys.2016.10.010
发表时间：
2017
期刊：
Comput. Environ. Urban Syst.
影响因子：
--
作者：
Chaowei Yang;Manzhu Yu;F. Hu;Yongyao Jiang;Yun Li
通讯作者：
Chaowei Yang;Manzhu Yu;F. Hu;Yongyao Jiang;Yun Li