The Domain Name System (DNS) is a ubiquitous part of everyday computing, translating human-friendly machine names to numeric IP addresses. With its redundant design, aggressive caching, and widely-assumed reliability, few suspect its internal failures as a source of delays. We show, through careful measurement, that the infrastructure responsible for resolving DNS names often encounters various failures which then induce delays. A systematic examination of the problem shows that the failures are widespread, uncorrelated, and can be a significant source of DNS-related delays.

We address this problem via the development of CoDNS, a cooperative DNS lookup service. It uses a locality and proximity-aware design to achieve low-latency, low-overhead name resolution in the presence of local DNS nameserver delay/failure. We show via repeated measurement and live traffic that CoDNS is an effective solution to DNS problems, and eliminates a major source of delay.