Internet Distance Prediction Using Node-Pair Geography

Abstract

Abstract-Predictive methods for learning network distances are often more desirable than direct performance measurements between end hosts. Yet, predicting network distances remains an open and difficult problem, as the results from a number of comparative and analytical studies have shown. From an application requirements perspective, there is significant room for improvement in achieving prediction accuracies at a satisfactory level. In this paper, we develop and analyze a new, machine learning-based approach to distance prediction that seeks to capture and generalize geographical characteristics between Internet node pairs, instead of relying on direct and ongoing measurements of partial paths. We apply a basic algorithm in machine learning to demonstrate this idea and highlight the potential benefits that this method may offer over other popular methods that exist today.

TNDP: Tensor-Based Network Distance Prediction With Confidence Intervals

“…The basic idea of these approaches is to map nodes into non-Euclidean space, for example, matrix factorization [12], [8], Hyperbolic embedding [34], [35], and network geography [14], and model the network distances in non-Euclidean curves. Generally, most of them can more accurately figure out network distances than Euclidean embedding since the predicted network distances are persistent in the real networks [8], [34].…”

“…The main idea is to exploit the limited network distances among a small set of node-pairs, in the form of either one-way delay or more often Round-Trip Time (RTT), to predict unknown distances among other node-pairs, where direct measurements are not performed. Over the past years, a large number of approaches have been developed [8], [7], [9], [10], [11], [12], [13], [14], [15] to predict network distance by introducing network coordinates. These approaches can be fell under two categories: Euclidean embedding and Non-Euclidean embedding (see more details in Section 5).…”

Docker Layer Placement for On-Demand Provisioning of Services on Edge Clouds

“…Edge clouds can download layers from other nearby clouds in the edge network so we do not have to fetch all necessary data from far away storage nodes in the backbone network. Research [32] on the distance between two servers shows that each 1000km adds 20 ms delay with a minimum threshold of 20ms RTT. As our edge clouds are all located nearby in the edge network, we model this latency as a uniform distribution between 10 and 30ms.…”

Network distance prediction for enabling service-oriented applications over large-scale networks

“…However, the existing researches have not taken full advantage of data-driven thought to steer the design of NDP. The quasi prototype of the datadriven NDP approaches can be found in [15], where an Internet NDP approach seeks to capture geographical characteristics between Internet host pairs by machine learning, instead of relying on direct measurements. Although without explicitly exposing data-driven thought, it still reveals a novel and efficient solution to the NDP.…”

“…For instance, in content delivery networks, an enduser can conveniently obtain its desired Web resources from a particular site with the knowledge of the predicted network distance. Similar to [2][3][4][5][6][7][8][9][10][11][12][13][14][15], in this article the network distance between two nodes is defined as the communication delay or latency between them, in the form of either one-way delay or more often Round Trip Time (RTT). Obviously, it is infeasible to ceaselessly probe network distances among all pairwise nodes in large-scale networks because global accurate measurements are difficult and costly to achieve and maintain.…”