当前位置: X-MOL 学术IEEE ACM Trans. Netw. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Optimizing in the Dark: Learning Optimal Network Resource Reservation Through a Simple Request Interface
IEEE/ACM Transactions on Networking ( IF 3.0 ) Pub Date : 2021-01-06 , DOI: 10.1109/tnet.2020.3045595
Qiao Xiang 1 , Haitao Yu 2 , James Aspnes 3 , Franck Le 4 , Chin Guok 5 , Linghe Kong 6 , Y. Richard Yang 3
Affiliation  

Network resource reservation systems are being developed and deployed, driven by the demand and substantial benefits of providing performance predictability for modern distributed applications. However, existing systems suffer limitations: They either are inefficient in finding the optimal resource reservation, or cause private information ( e.g. , from the network infrastructure) to be exposed ( e.g. , to the user). In this paper, we design BoxOpt, a novel system that leverages efficient oracle construction techniques in optimization and learning theory to automatically, and swiftly learn the optimal resource reservations without exchanging any private information between the network and the user. In BoxOpt, we first model the simple reservation interface adopted in most reservation systems as a resource membership oracle. Second, we develop an efficient algorithm that constructs a resource separation oracle by a linear number of calls on resource membership oracle. Third, we develop a generic framework to construct a resource optimization oracle by iteratively calling the resource separation oracle, and then develop three novel, efficient algorithms under this generic framework, the best of which computes the optimal resource reservation by a linear number of calls on resource separation oracle. As such, BoxOpt can discover the optimal resource reservation with $O(n^{2})$ calls on the resource membership oracle. We implement a prototype of BoxOpt with and demonstrate its efficiency and efficacy via extensive experiments using real network topology and a 7-day trace from a large operational federation network. Results show that (1) BoxOpt has a 100% correctness ratio by comparing with a state-of-the-art optimization solver, and (2) for 90% of requests, BoxOpt learns the optimal resource reservation within 10 seconds.

中文翻译:

在黑暗中进行优化:通过简单的请求界面学习最佳的网络资源预留

在为现代分布式应用程序提供性能可预测性的需求和实质利益的推动下,正在开发和部署网络资源保留系统。但是,现有系统存在局限性:它们要么无法有效地找到最佳的资源预留,要么会导致私人信息( 例如 ,从网络基础架构) 例如 ,给用户)。在本文中,我们设计了BoxOpt,这是一个新颖的系统,它利用优化和学习理论中的有效oracle构造技术来自动,快速地学习最佳资源预留,而无需在网络和用户之间交换任何私人信息。在BoxOpt中,我们首先将大多数预订系统中采用的简单预订接口建模为资源成员Oracle。其次,我们开发了一种有效的算法,该算法通过对资源隶属度oracle进行线性调用来构造资源分离oracle。第三,我们开发了一个通用框架,通过迭代调用资源分离oracle来构建资源优化oracle,然后在该通用框架下开发三种新颖,高效的算法,最好的方法是通过对资源分离oracle进行线性调用来计算最佳资源预留。因此,BoxOpt可以通过以下方式发现最佳的资源预留 $ O(n ^ {2})$ 调用资源成员oracle。我们实现了BoxOpt的原型,并通过使用真实网络拓扑的大型实验以及来自大型运营联盟网络的7天跟踪,展示了其效率和功效。结果表明(1)BoxOpt通过与最新的优化求解器进行比较,具有100%的正确率;(2)对于90%的请求,BoxOpt会在10秒内学习到最佳的资源预留。
更新日期:2021-01-06
down
wechat
bug