The Internet Report

This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. On today’s episode, our newest host and Technical Marketing Engineer, Chris Villemez, is joined by Kemal Sanjta, Principal Engineer, to dive into the details of the recent AWS outages from December 7th, 10th and 15th. They’ll walk through what ThousandEyes saw from its fleet of vantage points, as well as share some insight into what enterprises can learn from these incidents to build resilient cloud architectures.

00:00 Welcome: This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why.

00:15 Headlines: Today we’re going to do a thorough analysis of the major Facebook outage that took place yesterday, Monday, October 4. I’m joined by ​​Gustavo Ramos, ThousandEyes’ in-house expert on Network Engineering.

ThousandEyes Blog: https://www.thousandeyes.com/blog/facebook-outage-analysis

Analysis from Facebook: https://engineering.fb.com/2021/10/05/networking-traffic/outage-details/

1:17 Under the Hood: We'll walk through the sequence of events that led to this outage, understand what went wrong (and what actions may have made the situation worse), and what lessons we can all learn from this outage.

25:40 Outro: We've been on a bit of a break for the past several months as things were relatively quiet on the Internet front and for the foreseeable future we'll be a bit reactive in our episodes, when something major happens trust we'll be here. Questions? Feedback? Have an idea for a guest? Send us an email at [email protected]

00:00 Welcome: This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why.

00:08 Headlines: Today, Mike Hicks (Principal Solutions Analyst, ThousandEyes) and I discuss a recent BGP routing incident that had intermittent impacts on Amazon’s services, including Amazon.com and AWS compute resources, during a five-hour period on July 12.

01:04 Under the Hood: When we look into BGP routing at the time, we can see multiple BGP path changes due to a service provider erroneously inserting themselves into the path for a large number of Amazon routes. Watch this episode to see how the BGP incident led to significant packet loss, resulting in service disruption for some Amazon and AWS users. We also discuss why enterprises need to have continuous oversight of the paths their traffic takes over the Internet.

17:58 Outro: Questions? Feedback? Have an idea for a guest? Send us an email at [email protected]

This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. I’m joined today by Mike Hicks, principal solutions analyst here at ThousandEyes, to cover the outage of Akamai’s DNS service. The outage, which occurred on July 22nd around 3:38 PM UTC (8:38AM PT), struck during the course of business hours in Europe and North America, resulting in widespread impacts to applications and services hosted within Akamai servers. The outage itself was short-lived and was resolved roughly one hour after the outage began.

In this episode, we examine the customer impact, the relationship between DNS and CDNs, and what enterprises should take away from the incident. We also discuss the question of when it might make sense to invest in DNS or CDN redundancy—and when it is, frankly, overkill. Watch this week’s episode to hear our take, and as always let us know on Twitter what you think.

00:00 Welcome:This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why.

00:13 Headlines: Today, Kemal and I unpack an interesting BGP incident, in which a large-scale route leak briefly altered traffic patterns across the Internet.

00:58 Under the Hood: The incident began on Thursday, June 3rd at around 10:24 UTC, and resulted in a significant spike in packet loss that was noticeable in ThousandEyes tests. While this packet loss resolved within the hour (at around 10:48 UTC), we observed some interesting routing changes during this window—as traffic was diverted to a Russian telecom provider that had not previously been in the path. Watch this episode as we explore how this network provider managed to get itself into the routing paths of many major services, and why network visibility is so important to recognize these types of incidents in which your site may still be reachable but your traffic is being sent through an unexpected network.

20:45 Outro: Questions? Feedback? Have an idea for a guest? Send us an email at [email protected]

This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. I’m joined by ThousandEyes’ BGP expert, Kemal Sanjta, to review the June 16th outage of Prolexic Routed, a DDoS Mitigation Service operated by Akamai. According to a statement from Akamai, the outage was not due to a DDoS attack or system update, but instead a routing table limitation that was inadvertently exceeded.

In this episode, Kemal and I analyzed what happened and how customers of Akamai Prolexic who had automated failover mechanisms in place were able to recover more quickly than those that had to manually switch over to other providers. Watch this episode to learn more about this outage, and how different operational processes resulted in very different service outcomes.

00:00 Welcome: This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why.

00:12 Headlines: Today, I’m joined by Hans Ashlock, Director of Technology & Innovation at ThousandEyes, to unpack today’s major outage at Fastly, a popular CDN provider.

3:46 Under the Hood: Today, I’m joined by Hans Ashlock, Director of Technology & Innovation at ThousandEyes, to unpack today’s major outage at Fastly, a popular CDN provider. The widespread outage occurred around 9:50 UTC, about 5:50 am ET, and mostly impacted users across Europe and Asia due to the timing. he outage lasted approximately one hour until 10:50 UTC, yet residual impacts were felt beyond that. Today’s outage is a good example of the importance of having outside-in visibility not just across your app, but also to your app’s edge and all its dependent services.

39:05 Outro: Questions? Feedback? Have an idea for a guest? Send us an email at [email protected]

This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. I’m joined today by Mike Hicks, Principle Solution Analyst at ThousandEyes, to cover two recent application-related outages. The first occurred on May 19th around 12:50 UTC at Coinbase—a well-known cryptocurrency exchange. Around the time that news broke saying that the Chinese government would be imposing strict regulation on cryptocurrencies, users attempting to execute transactions were unable to access the application. From the ThousandEyes platform we were able to see a drop in availability around this time as well as increased load times (which in some cases resulted in timeout errors).

The second outage happened on May 20th around 17:35 UTC at Slack—an enterprise collaboration platform. While the outage was resolved within 90 minutes, it occurred during normal US business hours, making it particularly disruptive to users attempting to reach the application. These instances remind us that applications, much like the underlying networks they run on, can experience outages, and effective troubleshooting requires end-to-end visibility into both.

00:00 Welcome

00:14 Headlines: DNS and BGP and DDoS Attacks—Oh, My! This week we cover a couple of recent service degradation incidents involving DNS providers

2:19 Under the Hood: Kemal Sanjta, ThousandEyes’ resident BGP expert, joins us to discuss the May 6th disruption to Neustar’s UltraDNS service, which lasted nearly four hours. We discuss the BGP routing changes we observed during the incident and what they can tell us about the cause of the disruption. We also cover a separate incident involving Quad 9, a public recursive resolver service, which the company says was caused by a DDoS attack on May 3rd.

16:19 Expert Spotlight: Michael Batchelder (A.K.A., Binky), is here to discuss the two “Ds” of the Internet — DDoS attacks and the DNS Questions for Binky? Contact him at [email protected]

31:49 Outro: Questions? Feedback? Have an idea for a guest? Send us an email at [email protected]

This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. Today, we focused on an interesting outage that impacted Cloudflare Magic Transit, a relatively new offering from the CDN provider which aims to efficiently route and protect the network traffic of its customers. On May 3rd at approximately 3:00 PM PDT (10:00 PM UTC), ThousandEyes vantage points connecting to sites using Magic Transit began to detect significant packet loss at Cloudflare’s network edge—with the loss continuing at varying levels, for approximately 2 hours.

While the outage impacted some Magic Transit customers more significantly than others, we also observed mitigation actions by at least one customer to avoid the outage and restore the availability of their service to their users. This outage reminds us that no provider is immune to outages, even cloud and global CDN providers. However, with proactive visibility, you can respond quickly to reduce outage impact on your users. Watch this week’s episode to hear more about the outage from the ThousandEyes perspective.