To reduce the costs of repeated failures (e.g., from erroneously looping clients), we implemented two forms of negative caching. First, when a DNS lookup failure occurs, we cache the negative result for one hour. Second, when an object retrieval failure occurs, we cache the negative result for a parameterized period of time, settable via the negative_ttl variable in the cached.conf file. The default value is 5 minutes.
The cache also implements a cache of successful DNS lookups. The timeout for this cache is ``hard-coded'' at one day.