Up-to-date latency information as provided by a latency service is crucial for many distributed applications. Network coordinates are an efficient and scalable mechanism for obtaining latency estimates. However, any practical implementation must handle the variance of latency samples and minimize measurement overhead, while ensuring stable and accurate coordinates. In this paper, we have described the APIs of a network coordinate service and a library. We have also shown how statistical filtering addresses sample variance, how the distinction between system- and application-level coordinates improves coordinate stability, and how the use of application-level traffic for coordinate updates can reduce overhead. We believe that a network coordinate service can add network awareness to a wide range of applications and become one of a set of standard services for planetary-scale applications.
This material is based upon work supported by the National Science Foundation under Grant No. 0330244.
