This page has all the materials that we used in the work Measuring Update Performance and Consistency Anomalies in Managed DNS Services including the trace we collected for the measurement of lookup, update, consistency.
Some supplementary experiments that we didn't mention in the submission are also included.
We attach our interview questions at the end.
For each service provider, we measure 500 lookup latencies from across 40 PlanetLab nodes (over 20,000 data points for each service provider). We ensure that lookups are served from the name servers maintained by its provider by requesting the name servers through their external IP anycast addresses directly. We apply a 1 second trim filter to rule out the long tail latency. We show the lookup latency in the Figure below. Here is the link to download the trace for lookup latency.
Our result is consistent with the result by DNSPerf.
Here is the link to download the trace for update propagation latency among authoritative name servers..
We use 1,000 RIPE probes and only 961 are responsive. The queries from 8 probes are always timed out. Totally, we get result from 953 probes.
Here is the link to download the trace collected with 1000 RIPE probes...
March and April: the link to download the trace of March and April, 2018..
May: the link to download the trace of May, 2018.
June: the link to download the trace of June, 2018
July: the link to download the trace of July, 2018
August: the link to download the trace of August, 2018
September: the link to download the trace of September, 2018
October: the link to download the trace of October, 2018
November: the link to download the trace of November, 2018
| ID | Question |
|---|---|
| Q1 | What is the major difference between your DNS service and the others? |
| Q2 | Do you use anycast network to improve your performance? How many anycast network locations do you have? |
| Q3 | What is your expected lookup latency? |
| Q4 | Can I set my TTL to 0 by using your DNS service? |
| Q5 | What is your expected update latency? |
| Q6 | Do you support DNSSEC? Is it easy to set it up on your service? |
| Q7 | What security mechanism(s) do you use to prevent DDoS attack? |
| Q8 | Do you support an API to update DNS records managed by your service? |
| Q9 | What load balance schemes do you have? |
The answers are summarized based on the answer we got from the interview with the representative of the companies as well as their online document.
An interesting pricing comparison among top managed DNS providers done by DNSMadeEasy can be found here: https://dnsmadeeasy.com/pricing/pricecomparison/
A service comparison among top managed DNS providers done by DNSMadeEasy can be found here: https://dnsmadeeasy.com/pricing/providercomparison/
| Question | UltraDNS | Verisign | NS1 | Akamai | Cloudflare |
|---|---|---|---|---|---|
| Q1 | Performance 100% uptime DNS shield DDoS attack protection |
100% uptime DDoS protection Optimize performance Traffic management |
Anycast network Cloud-based architecture Traffic filter |
Cloud-based DNS Improved performance High availability DDoS protection |
The fastest DNS provider DDoS protection Easy management |
| Q2 | 30 | 17 | 25 | ~100 | 151 |
| Q3 | Depends |
Depends |
Depends |
Depends |
12ms |
| Q4 | Yes | Yes | Yes | Not sure | No |
| Q5 | a few seconds | a few seconds | a few seconds | a few seconds | 5 seconds[1] |
| Q6 | Yes | Yes | Yes | Yes | Yes |
| Q7 | Cloud-based On-Premise Hybrid[2] |
OpenHbrid Cloud-based[3] |
Unknown |
Kona Site Defender 210,000 servers over 120 countries Web Application Firewall[4] |
Multilayer protection[5] |
| Q8 | Yes, with a business package |
Yes, with a business contract |
Yes |
Yes, with a business contract |
Yes |
| Q9 | Weighted-based Failover |
Lua script Weighted-based Failover Geo-location |
Chain Filter Geographic-based Weighted-based More options |
Mapping-based |
Geographic-based Failover |