APNIC’s PING Podcast Deep Dives into DNS Query Repetition, Transport Protocols, and the Cost of "Saying No"

In the latest episode of the PING podcast, APNIC Chief Scientist Geoff Huston joins the host for an in-depth continuation of their ongoing technical exploration into the mechanics of the Domain Name System (DNS). Specifically, the conversation focuses on the phenomenon of repeat queries observed across APNIC Labs’ experimental infrastructure, examining how evolving DNS protocol usage, modern record types, and different methods of handling negative responses impact overall network efficiency and server workload.

The historical context of these repeat queries has evolved significantly over time. As the architecture of the DNS has expanded to incorporate entirely new types of records that simply did not exist when many foundational internet protocols and troubleshooting methods were first deployed, the nature of network traffic has shifted. Today, systems frequently encounter scenarios where a single logical resolution process generates multiple queries targeting a variety of distinct record types and namespaces. This compounding effect has transformed how recursive resolvers, intermediate infrastructure, and authoritative servers interact, bringing new visibility to inefficiencies that were previously less pronounced at a global scale.

Central to the discussion is Huston’s ongoing research into two critical and interrelated aspects of DNS operations: the precise operational consequences of how servers choose to answer negative requests, and the systemic impacts of alternative transport protocols on query repetition and overall query load.

The Complex Economics of Saying "No" in the DNS

When a DNS query asks for a record that does not exist, or requests information that an authoritative server cannot or will not provide, the system must deliver a negative response. However, Huston’s exploration reveals that the mechanics of how different servers choose to communicate this "no" can vary wildly, and the resulting operational outcomes are frequently surprising.

Different strategies for saying no—ranging from explicit negative caching responses and standard error codes to the drastic approach of dropping the packet and not responding at all—produce vastly different outcomes across the network. These choices directly dictate the amount of work required not just by the authoritative servers at the edge, but also by intermediate recursive resolvers and the originating clients themselves. When a server fails to respond, for instance, it often triggers timeout behaviors and retransmission logic in upstream resolvers, paradoxically multiplying the query volume rather than suppressing it. By analyzing these variances, Huston’s research highlights how seemingly minor configuration choices in negative response handling can have outsized ripple effects on global DNS traffic volumes and infrastructure strain.

Transport Protocols and the Debate Over State

Beyond the challenges of negative responses, the PING podcast episode dives deeply into the role that alternative DNS transport protocols play in query repetition and total query load. For decades, the design of the DNS was heavily optimized around the User Datagram Protocol (UDP). Because UDP is connectionless, it avoids the overhead of establishing and maintaining explicit connection states, making it historically favored for high-volume, low-latency transaction models.

However, this reliance on UDP and its inherent behaviors has also been implicated as a contributing factor to the persistence of unnecessary repeated queries. For many years, network architects and protocol designers argued that the operational overhead of maintaining connection states for protocols like TCP—or modern connection-oriented alternatives—made them entirely unsuitable for the massive scale of the global DNS.

Yet, as the volume of unnecessary, repeated queries continues to scale upward across the internet, the calculus surrounding protocol efficiency is beginning to shift. Huston points out that as handling repeated, stateless UDP queries consumes increasing amounts of cumulative processing power and bandwidth, the end-to-end reality may be changing. It is becoming increasingly plausible that maintaining an explicit protocol state could actually require less total work over the lifecycle of a transaction than repeatedly handling the same redundant, stateless queries under heavy load conditions. This realization opens up broader industry conversations about the future evolution of DNS transport layers and how core internet protocols can better adapt to modern congestion and query amplification challenges.

As the technical community continues to grapple with these underlying architectural pressures, researchers and network operators are encouraged to engage with the findings. Readers interested in a deeper dive into the specific mechanics and data surrounding the problems of negative responses can find further documentation and analysis published directly on the APNIC Blog.

Listeners wishing to stay updated on future technical discussions can stream and subscribe to the PING podcast across standard distribution channels. APNIC continues to welcome contributions, insights, and research from the broader internet community. Professionals interested in sharing their own findings or providing feedback on the podcast and the APNIC Blog are invited to reach out directly to the editorial team to help drive ongoing improvements to community resources.

The views and opinions expressed in this technical discussion and associated research are those of the individual authors and do not necessarily reflect the official policy or position of APNIC. All participants and readers are reminded that the APNIC Blog operates under a strict Code of Conduct designed to foster a professional, collaborative environment.