29 September 2017 to 3 October 2017
Fairmont San Jose
US/Pacific timezone

Performance of ALIAS records vs CNAME

29 Sept 2017, 17:30
30m
Regency 2 Ballroom (Fairmont San Jose)

Regency 2 Ballroom

Fairmont San Jose

170 S Market Street, San Jose, 95113, CA, USA
Standard Presentation Public Workshop Public Workshop

Speaker

Mr Christopher Baker (Dyn)

Description

The Canonical Name, CNAME, record has become the default means of service integration for a number of Cloud and SaaS providers. The scope of services integrated via CNAME includes everything from marketing automation services to cloud load balancers. In some cases, you may have a service integration which is done by a CNAME and points to another CNAME, which may point to yet another CNAME. Some authoritative providers have implemented custom record types that seek to collapse CNAMEs to reduce the latency of the chain of DNS queries required to unwind such a CNAME chain. This presentation is an overview of recent testing of the Dyn+Oracle ALIAS record in the wild. We surveyed ASNs for those with the highest volume to our authoritative, then looked for RIPE Atlas probes in those networks. Probes were configured to use their default resolvers and execute two tests. In the first test, they requested a domain which was the start of a CNAME chain. In the second, they requested a domain associated with an ALIAS record, collapsing the same CNAME chain. This presentation contrasts the performance of the ALIAS record vs. the CNAME chain, as well as provides an overview of edge cases observed. This latter point is included to highlight the potential importance of a "fallback record" in the implementation of the ANAME ( https://tools.ietf.org/html/draft-ietf-dnsop-aname-00 ), a new record type intended to standardize this behavior across providers.
Talk Duration 30 Minutes

Primary author

Presentation materials