Self-healing routing: failure, modeling and analysis

Abstract

Network failures occur frequently, and self-healing ability of existing routing protocols cannot guarantee fast route convergence under these failures without impacting packet forwarding. During routing convergence, network routes may be incorrect and even routing black holes and loops occur, which will result in extensive packet loss and thus influence network performance. To solve this problem, several improved routing solutions have been proposed. In this paper, we propose the concept and model of self-healing routing and analyze the key problems in current intra-domain and inter-domain self-healing routing protocols after briefly reviewing the characteristics of network failures. We classify the different self-healing solutions into two categories based on their design principles: routing restoration and routing protection, and systematically analyze these different typical solutions. Finally, we discuss several key issues in self-healing routing and propose a hybrid protection and restoration based routing scheme.

More efficient automatic repair of large-scale programs using weak recompilation

“…Automatic program repair is considered to be more difficult and tedious than bug finding [1], and plays an important part on providing fully automated debugging. Recently some promising techniques [2][3][4][5][6][7] have been proposed to automate this process, reducing the cost of software maintenance. Generally, the repair process is three-phase procedure: locating the program bug (fault localization); generating the candidate patches based on some specified rules (patch generation); and validating these patches through various test cases over and over until a valid patch is found (patch validation).…”

Finding the best answer: measuring the optimization of public and authoritative DNS

“…On account of a failure of CDN nodes or routing jitter [6], all requests of different measurement nodes may route to the same destination. To reduce this bias, each domain and IP are measured at least 3 times over different time periods.…”