Attaching geographical locations to IP addresses in the context of a world-wide disaster monitoring system Hsueh-Hao Chang (hsuehhac) and Chi Zhang (chiz) In this project, we aim to develop a disaster monitoring system. We will design and implement a program which will periodically send request to some selected web servers to see their alive status. According to these response, the program can predict some bad things like earthquake or large-scale power failure might happen. In order to detect the behavior of web servers are abnormal, we need to measure these web servers when they are in normal condition. Some criteria might include mean and variance of RTT. This program might draw the affected area in the Google map for visualization. From the research point of view, we are expecting to invest into following issues: 1. How to select the most representative web servers and how to know their geographical locations group them together based on that. 2. Which method do we adopt to detect server aliveness? (e.g., ping?) During probing, how to detect and tolerate transient failure? We are also considering define a performance scoring system to show the severity of the disaster based on the connection stats. 3. How to set the right granularity in order to show geographical adjacency, neither too big nor too small? Where to add into geo-information to group a set of physically contiguous IP addresses? 4. In which way should we organize physical location information, e.g., flat or hierarchical? The choice of any manner of combining physical and IP information, t pose too much of maintenance overhead. E.g. IP address, in theory could be assign to web instance at any location. How to retrieve the right physical location upon bootstrapping? We will use simulation as the major way to validate our system since real-world disaster is unpredictable. We expect our system should correctly detect the abnormal behaviors of the servers which are affected by the disaster in the simulation. The system can also record the historic log.