Research question: Is DNS over CoAP meaningful for service discovery in edge K3s clusters?

[tang22164]

I am not reporting a bug or requesting a feature implementation.
I would like to ask whether this research direction and use case are technically reasonable from the K3s/edge Kubernetes perspective.

Background

I am a student in Japan, and I am considering a graduation research project related to service discovery in edge K3s clusters.

In my research idea, IoT devices communicate with services running in an edge environment.
The edge environment is assumed to run a lightweight Kubernetes distribution such as K3s.

In a normal Kubernetes/K3s environment, CoreDNS is used for service discovery inside the cluster.
For example, Pods can resolve Service names through DNS and communicate with other Pods or Services.

I am interested in whether DNS over CoAP could be used instead of, or alongside, the standard CoreDNS-based service discovery mechanism in an edge K3s environment.

Research idea

The basic idea is to implement a DNS over CoAP mechanism for an edge K3s cluster and compare it with the standard CoreDNS-based approach.

For example, the system could work as follows:

• A client sends a DNS query over CoAP.
• A DNS over CoAP gateway or resolver receives the request.
• The gateway either forwards the query to CoreDNS or obtains Kubernetes Service and Endpoint information.
• The gateway returns the DNS response or service information to the client over CoAP.

I am considering two possible designs:

1. Using DNS over CoAP for service discovery inside the K3s cluster.
2. Using DNS over CoAP only between IoT devices and the edge gateway, while keeping normal CoreDNS-based DNS inside the K3s cluster.

The goal is to evaluate whether DNS over CoAP can reduce latency, communication overhead, CPU usage, memory usage, or power consumption compared with the standard CoreDNS-based approach.

Motivation

CoAP is designed for constrained IoT devices and low-power networks.
Therefore, I thought DNS over CoAP might be useful in edge computing environments where IoT devices communicate with services running on a nearby K3s cluster.

However, I am not sure whether applying DNS over CoAP inside the K3s cluster itself is meaningful.

Communication inside a K3s cluster usually happens within the same node or over a local network.
It is not usually a low-power wireless network like the communication between IoT devices and an edge gateway.

Because of this, I suspect that replacing or bypassing normal CoreDNS-based DNS with DNS over CoAP inside the K3s cluster may not significantly reduce power consumption or improve latency.

Concern

My main concern is that the performance difference may be too small to measure clearly.

For example, if the communication is only inside the K3s cluster, the difference between normal DNS and DNS over CoAP may be much smaller than other factors such as:

• application processing time
• container networking overhead
• Kubernetes Service routing
• CPU load on the edge node
• cache behavior
• network interface power consumption

Therefore, even if DNS over CoAP has smaller protocol overhead in some cases, it may not lead to a meaningful reduction in total power consumption.

I am also concerned that introducing DNS over CoAP may reduce compatibility with the standard Kubernetes service discovery model, unless it is implemented as a gateway in front of CoreDNS.

Understanding of DNS over CoAP

I am still learning about DNS over CoAP, so I would also appreciate comments on whether this is the correct term and approach for my idea.

My current understanding is that DNS over CoAP means carrying DNS queries and responses using CoAP as the transport/application protocol.

I am not trying to request a new K3s or CoreDNS feature at this stage.
Rather, I would like to understand whether DNS over CoAP is a reasonable research direction for edge K3s service discovery, or whether this idea should be considered outside the scope of K3s/CoreDNS/Kubernetes.

Community and use case question

I would also like to ask whether using DNS over CoAP is considered a reasonable direction from the K3s/edge Kubernetes community perspective.

In particular, I would like to know whether such a use case is realistic in edge Kubernetes environments.

For example:

• Are there real-world situations where DNS or service discovery overhead becomes a problem in edge K3s clusters?
• Are there IoT or edge scenarios where DNS over CoAP would be more appropriate than normal DNS/CoreDNS?
• Would the community consider this kind of approach useful, experimental, or mostly out of scope for K3s/Kubernetes?
• Is DNS over CoAP better suited for communication between IoT devices and the edge gateway, rather than for communication inside the K3s cluster?

Research question

Is DNS over CoAP meaningful for service discovery in an edge K3s cluster?

More specifically, I would like to know whether it makes sense to apply DNS over CoAP inside the K3s cluster, or whether it should be used only between IoT devices and the edge gateway.

Possible comparison targets

I am considering comparing the following approaches:

• Standard CoreDNS-based DNS service discovery
• DNS over CoAP gateway
• HTTP-based service discovery gateway
• Static service configuration

Evaluation metrics

I am considering the following evaluation metrics:

• Name resolution latency
• Packet size
• Total communication overhead
• CPU usage
• Memory usage
• Power consumption of the edge node
• Power consumption of the IoT device, if an actual device is used
• Implementation complexity
• Compatibility with Kubernetes/K3s

Questions

1. From the K3s/CoreDNS perspective, is DNS over CoAP a technically reasonable research direction for service discovery in edge K3s clusters?

2. Are there realistic use cases where DNS or service discovery overhead becomes a problem in edge K3s environments?

3. Is DNS over CoAP likely to provide measurable benefits inside a K3s cluster, such as lower latency, lower communication overhead, or lower power consumption?

4. Would DNS over CoAP be more suitable between IoT devices and the edge gateway, while keeping the normal CoreDNS-based DNS mechanism inside the K3s cluster?

5. Are there existing metrics, benchmarks, or related projects that would be useful for evaluating CoreDNS/K3s service discovery performance?

I would appreciate any comments on whether my technical assumptions are reasonable, whether such a use case exists, and whether this research direction seems meaningful.

Thank you.

[miri64]

I am still learning about DNS over CoAP, so I would also appreciate comments on whether this is the correct term and approach for my idea.

I am a co-author of RFC9953. If you have any questions: feel free to ask! :-)