Unraveling the Mechanics of DNS Repeat Queries: Insights from APNIC Labs

The intricacies of the Domain Name System (DNS) continue to present evolving challenges for network operators, researchers, and protocol designers alike. In the latest episode of the PING podcast, APNIC Chief Scientist Geoff Huston joins the conversation to examine a persistent phenomenon observed within APNIC Labs experiments: the continuous stream of repeat queries hitting the infrastructure. Over time, the nature and volume of these redundant requests have shifted significantly, driven largely by the ongoing evolution of DNS protocol usage, the emergence of entirely new record types that simply did not exist when these architectural patterns first surfaced, and the complex ways modern systems attempt to resolve missing or negative answers.

This deep dive into DNS behavior sheds light on how the foundational infrastructure of the internet handles the unknown, and why the choices made when saying "no" to a query can have profound ripple effects across the global network. As internet protocols and infrastructure scale to meet modern demands, understanding the exact cost of these repeat queries has become more critical than ever.

Exploring the Cost and Consequences of Saying "No"

At the heart of Geoff Huston’s recent investigations for APNIC Labs is a fundamental question: how do different methods of answering "no" within the DNS architecture impact overall query volume and resource utilization? When a resolver asks for a domain or a record that does not exist, the authoritative server, intermediate resolvers, and the originating client must all process that negative response. However, the mechanics of how that negative response is delivered—or whether it is delivered at all—vary widely.

Huston’s research reveals that the various ways of communicating a negative outcome can yield surprisingly divergent results. This includes scenarios where authoritative servers or resolvers choose not to respond at all. Rather than simply silencing unnecessary traffic, dropping packets or altering standard negative caching behaviors can inadvertently trigger secondary behaviors in client applications and resolvers. These systems often retry failed lookups aggressively, multiplying the original query volume and placing an unexpected burden on authoritative servers and intermediate network nodes alike.

The resulting traffic amplification means that a single missing resource record can generate a cascading storm of repeat queries. This dynamic not only consumes valuable processing cycles at the authoritative layer but also contributes to unnecessary background noise across the wider internet. By analyzing these patterns, Huston aims to map out the exact trade-offs associated with different negative response strategies, offering network administrators a clearer picture of how protocol configurations influence overall network health.

Re-evaluating Transport Protocols and State Maintenance in DNS

Beyond the mechanics of negative responses, Huston’s ongoing research turns a critical eye toward the transport protocols used to carry DNS traffic. For decades, the Domain Name System has relied heavily on the User Datagram Protocol (UDP) for standard queries, prized for its low overhead and connectionless nature. Because UDP does not require the establishment or maintenance of a persistent connection state, it has long been considered the ideal vehicle for handling high-volume, stateless transactions at scale.

However, the proliferation of unnecessary and repeated queries is forcing researchers to re-examine this long-standing assumption. As the sheer volume of redundant traffic continues to climb, the cumulative cost of processing millions of independent, stateless UDP transactions is beginning to stack up. Many architects and protocol designers historically argued that the overhead associated with maintaining connection states—such as the handshake and state management required by TCP and modern transport alternatives—made them entirely unsuitable for the sheer scale of the global DNS.

Yet, as Huston points out, the calculus may be shifting. When systems are constantly bombarded with high volumes of repeated queries that yield no new information, the lack of state in traditional UDP-based exchanges can paradoxically lead to massive inefficiencies. End-to-end, maintaining a persistent protocol state might ultimately require fewer system resources and less cumulative work than repeatedly handling a relentless stream of identical, disconnected queries. This provocative insight challenges conventional wisdom regarding transport layer selection for core internet infrastructure and opens up new avenues for discussion around future protocol optimizations.

Continuing the Conversation and Community Engagement

The exploration of DNS repeat queries, negative response handling, and transport protocol efficiency forms a vital part of the ongoing dialogue within the technical community. As APNIC continues to track these trends through its continuous laboratory experiments, the findings offer valuable empirical data for engineers working to keep the internet resilient, scalable, and performant.

The discussion captured in this episode of PING serves as a reminder of how deeply interconnected protocol design choices are with real-world network operations. From the way servers handle empty answers to the ongoing debate over stateful versus stateless transport mechanisms, the minutiae of DNS engineering directly impacts global internet traffic patterns.

Listeners interested in delving deeper into the nuances of negative DNS responses can find additional commentary and detailed technical write-ups on the APNIC Blog. The PING podcast remains accessible across major streaming platforms, offering regular updates and discussions on critical internet routing, addressing, and protocol developments.

APNIC continues to encourage researchers, network operators, and community members to share their own insights, data, and stories. Those with relevant research or unique operational perspectives are invited to reach out directly to the production team to contribute to future episodes. Feedback on both the podcast and the APNIC Blog is always welcomed to ensure the content continues to support and inform the broader internet community.