IBM NS1 Connect

 View Only

How to Read and Compare DNS Performance Statistics

By Ben Ball posted Wed September 27, 2023 11:58 AM


Fast DNS is important. Naturally, every business wants to deliver applications, services, and content to users with the best possible performance. So when they shop around for an authoritative DNS service, it’s only natural that businesses want to compare the response speeds of different providers.

There are many sources of performance data out there which will give you a sense of how various DNS providers stack up against one another. DNSperf is a common source of performance data, and Catchpoint is another. Everyone has their favorite.

It’s tempting to look at these raw statistics and draw definitive conclusions about which service is “better”. Yet it’s important to know both the context of how these statistics are created and how to interpret them in an operational setting. Adding a little bit of nuance to the numbers can help businesses make more informed decisions.

Here are a few points to consider when benchmarking DNS performance:

Most publicly available DNS performance data is synthetic

The publicly available DNS performance statistics you find on the internet aren’t derived from Real User Monitoring (RUM) data sources. They’re created by generating DNS queries from a representative sample of top tier data centers and pinging different services. The results of those queries are then averaged out and assumed to represent actual performance.

In reality, you’d need access to a broad spectrum of user data and information from the DNS providers themselves to generate truly accurate latency information. Given the significant variation in internet traffic patterns around the world, no sampling methodology is going to provide anything more than a general point of comparison when it comes to DNS query performance.

Latency correlates with geography more than network performance

Geography is the most significant factor in the performance of any DNS network. If the DNS provider has a nearby Point of Presence (PoP), the response latency will be lower. It’s just physics in the end - response times are mostly dictated by the speed of light (or current).

This is why DNS performance statistics are generally better for providers with more PoPs. Some sites measure latency of DNS queries around the world, naturally favoring the few companies with the resources and customer base to deliver consistently high performance everywhere. North America (or US-only) statistics tend to be more forgiving, as most providers can get away with just a few regional PoPs to claim parity with top tier providers.

With all of this in mind, it’s important to ask where your DNS traffic is likely to come from. A North America-only business will read the latency numbers differently from a truly global business.

Specific millisecond latency rates don’t matter

Will your users be able to tell the difference between a query returned in 5ms and a query returned in 10ms? Probably not. They’re both really fast. Even when you’re doing multiple DNS lookups at a time, the performance difference is almost undetectable.

It’s sort of like comparing high-definition TVs: at a certain level, the human eye can’t tell the difference, even when the frame rate is faster. Diminishing returns set in.

DNS performance matters, but it can only be measured in rough categories. If query responses come back under 20ms or so, you’re doing well. Nobody will notice if one provider is “better” than another under that number. They’re all roughly equivalent.

The same is generally true even when latency numbers increase. A 50ms response is worse than a 20ms response, but it’s comparable to a 70ms response. The relative values are more important than the actual values.

Not all DNS services are really DNS services

Another question to ask when you’re comparing DNS performance statistics: “Do I associate this company with DNS services?” There are plenty of DNS providers who appear in the performance statistics but don’t actually offer DNS services.

Domain registrars and build-your-own website providers like WordPress fall into this category. Both will respond to DNS queries on your behalf, but they will not offer you the level of expertise and DNS redundancy that NS1 or any other specialist providers can. You can only buy DNS from them if you’re a customer of their other services.

ISPs and other default DNS providers also fall into this bucket. They run DNS on behalf of their customers, but it’s not a separate offering that you can buy. In that sense, they aren’t really comparable as a true dedicated DNS product.

Key Takeaways for Comparing DNS Performance

All this to say that DNS performance statistics can be generally useful as a point of comparison, but it’s important to know both the context of the numbers and how they apply to your business needs. While it’s tempting to look at the stacked ranking of providers and simply go with the fastest one, there’s a ton of meaningful detail underneath those numbers that shouldn’t be discounted.

In the end, it’s important to go with an authoritative DNS provider which meets your operational needs. That’s always going to be a combination of DNS performance and something else - functionality, customer service, price, experience. Performance is useful and important, but it’s not the only important factor.