By using CoDNS, CoDeeN obtains other benefits in capacity and availability, and these may apply to other applications as well. The capacity improvements come from CoDeeN being able to use nodes that are virtually unusable due to local DNS problems. At any given time, roughly 10 of the 100 PlanetLab nodes that run CoDeeN are experiencing significant DNS problems, ranging from high failure rates to complete failure of the primary (and even secondary) nameservers. CoDeeN nodes normally report their local status to each other, and before CoDNS, these nodes would tell other nodes to avoid them due to the DNS problems. With CoDNS, these nodes can still be used, providing an additional 10% extra capacity.
The availability improvements come from reducing startup time, which can be dramatic on some nodes. CoDeeN software upgrades are not announced downtimes, because on nodes with working local DNS, CoDeeN normally starts in 10-15 seconds. This startup process is fast enough that few people notice a service disruption. Part of this time is spent in resolving the names of all CoDeeN peers, and when the primary DNS server is failing, each lookup normally requires over five seconds. For 120 peers, this raises the startup time to over 10 minutes, which is a noticeable service outage. If CoDNS is already running on the node, startup times are virtually unaffected by local failure, since CoDNS is already sending all queries to remote servers in this environment. If CoDNS starts concurrently with CoDeeN, the startup time for CoDeeN is roughly 20 seconds.