Product News
Announcing Cloud Insights for Amazon Web Services

The Internet Report

Twitter to X: Charting Performance and Outages

By Mike Hicks
| | 12 min read
Internet Report on Apple Podcasts Internet Report on Spotify Internet Report on SoundCloud

Summary

The 18 months from sale to domain transition were eventful, but appeared to help ensure a smooth shift from twitter.com to x.com.


This is The Internet Report, where we analyze recent outages and trends across the Internet, through the lens of ThousandEyes Internet and Cloud Intelligence. This week, we’re taking a break from our regular programming to go in-depth on the Twitter transition, which now appears to be complete. We will resume our usual blog and podcast in the next episode.


Tens of thousands of mergers and acquisitions (M&As) are announced each year, but one deal in late 2022 has dominated discussions since it was struck: the $44 billion purchase of Twitter by entrepreneur Elon Musk.

A key reason for this interest is that the platform hosted hundreds of millions of daily active users before the sale. Would the years of development work put into the platform and its strong track record of application performance be preserved?

To be fair, Twitter (and its users) had seen and experienced outages before its purchase, although they were relatively sporadic. For example, Twitter once crashed during a Steve Jobs keynote in 2008, and up until 2012, its popularity in news events sometimes led to network capacity issues—during which time an iconic "fail whale" illustration was displayed on the homepage. Additionally, in January 2016, Twitter experienced a sizable international outage.

Some outages weren’t Twitter’s doing: for example, a BGP hijack in early 2022 where an ISP erroneously advertised its network as the best way to get to Twitter, with the error propagating across much of Europe before it could be corrected.

Still, an eight-year gap between major incidents for a platform of Twitter’s size, scale, and resources presented an enviable record. It spoke to the resilience of the architecture—one that had been designed with scale in mind and had been fine-tuned and tweaked over the years.

It appeared that the architecture was well-built for its purpose and scale. In essence, the platform seemed to have a strong foundation. However, following the acquisition, ThousandEyes noticed a rise in the frequency of what seemed to be application issues and disruptions. This trend continued until May 17, 2024, when the full migration to X.com was completed. Since that date, there doesn't appear to have been any significant disruption to the X.com platform, with the latest disruption of any note taking place on July 1 (more on that incident below).

The Sale and Post-sale Periods

Typically leading into an M&A, and then in the months following completion, we’d expect to see relative stability of IT and network operations. The efficacy of technology systems is likely to be tested to some degree during due diligence; and companies typically buy companies they don’t need to put a whole lot of work into in order to realize synergies and generate value. Eventually, however, some work is required to consolidate both sides’ technology environments, and during that burst of activity, one might anticipate seeing some external-facing impacts.

Twitter appears to be an outlier in that respect. Instead of displaying the usual post-acquisition stability, ThousandEyes’ data indicates that the number and frequency of outages did increase commensurate with the acquisition. If we compare the two periods (pre- and post-acquisition), it’s apparent that there were fewer and less frequent outages prior to the purchase.

Graph comparing Twitter outage counts before and after acquisition.
Figure 1. Comparison of outage counts before and after acquisition

Almost immediately after the acquisition, ThousandEyes observed more frequent, short-lived backend issues that were causing 5xx and 4xx errors for users; a redirect loop leading to slow performance and timeouts; and unresponsive services affecting functionality, such as the delivery of SMS codes used in two-factor authentication.

The situation reached an early crescendo in November 2022, when there were concerns that the application might not survive a weekend, due to organizational restructuring. However, Twitter managed to overcome these challenges, and there has been no major outage since then. This, if nothing else, shows the strength of its architecture and the years of engineering work invested in its infrastructure and core systems.

And yet, this does not mean that Twitter has been incident-free over the last couple years.

For example, on the morning of March 1, 2023, Twitter users saw an empty timeline and a “Welcome to Twitter” message, as if they had just set up an account and logged in for the first time. While the site itself appeared to be functional, with trending topics working and users able to tweet, the timeline, which normally displays recent tweets from followed accounts, did not render properly and was blank. This incident affected morning users in the U.S. and lasted several hours. It was a departure from the typical backend-related outages that Twitter had previously experienced, which were largely made up of timeouts and/or unresponsive services. It's not entirely clear what caused this issue, but five days later, an internal system change caused broader disruption.

Since the beginning of 2024, Twitter has been encountering sporadic service disruptions. These interruptions were generally short lived, and impacted particular features of the service, rather than leading to complete unavailability of the app.

Graph of X outages observed in 2024 so far
Figure 2. A number of sporadic X outages observed in 2024 so far, many of them brief application issues

One of the most recent disruptions observed by ThousandEyes was in late April 2024, which manifested as application timeouts, indicating an issue with the service’s backend.

Domain Migration: The Culmination

Twitter marked a major milestone on May 17 when it effectively ceased to exist, with all core systems now running on the x.com domain. The transition appeared to go relatively smoothly—possibly due to preparatory work performed in the 18-month period since acquisition. It’s possible that some of these preparations could have manifested as degradations or even as outages—but they may have been important in the context of a domain migration goal to help ensure that went smoothly.

ThousandEyes was able to observe a castellation effect on traffic during the domain migration. We saw some 5xx errors, such as 503 Bad Gateway, and server timeouts. It seems that the HTTP 5xx error may be due to DNS settings being moved to point to a new name server.

ThousandEyes platform screenshot showing HTTP 5XX and server timeout errors observed during official domain change to X.com
Figure 3. HTTP 5XX and server timeout errors were observed during official domain change to X.com

This could potentially be caused by old indexed files, leading to HTTP requests getting stuck in a redirect loop and failing. Despite this, it doesn't appear to have any significant impact on the user experience.

After the domain cutover, things appeared to be operating without a glitch and have continued to do so overall.


Explore the domain migration further in the ThousandEyes platform (no login required).

Screenshot from ThousandEyes platform showing that post-domain change performance appeared to be normal.
Figure 4. Post domain change performance appeared to be normal

As mentioned above, the only incident of any note since the domain migration was a brief disruption on July 1.

At around 10:20 PM (UTC) on July 1, some users experienced issues accessing the service. While the application remained reachable from a network perspective, ThousandEyes data indicated that some requests from different regions were timing out when attempting to retrieve information from backend systems, suggesting a backend service issue. The disruption itself appeared relatively short, with most services appearing to be restored around 10:30 PM (UTC).

Screenshot from ThousandEyes platform showing requests to X.com services timing out for some regions
Figure 5. Requests to X.com services timing out for some regions

We’ll continue our observations as the platform evolves in its new X era.

By the Numbers

Transitioning from our discussion of X, let’s close by taking a look at some of the global trends ThousandEyes observed across ISPs, cloud service provider networks, collaboration app networks, and edge networks over three recent weeks (June 17 - July 7):

  • During the observed period, global outages initially increased but then began to decrease. From June 17-23, there was a 38% increase in outages compared to the previous week, rising from 186 to 256. However, the following week (June 24-30) saw an 18% decrease in the number of outages. This downward trend continued into the start of July, with outages dropping from 216 to 151 between July 1 and 7, marking a 30% decrease compared to the previous week.

  • The United States experienced a similar pattern: a 38% increase in outages from June 17-23, followed by a 29% decrease the next week (June 24-30). This was then followed by a further decrease of 21% from July 1-7. The decrease in both global and U.S. outages observed during this time is somewhat expected and consistent with ThousandEyes' typical observations for this time of year.

  • Only 34% of network outages occurred in the United States during the two-week period from June 17 to June 30. This is a decrease compared to the previous two-week period (June 3-16), during which outages in the U.S. represented 40% of all global outages.

  • Looking at the month-over-month data, in June, there were 890 outages observed worldwide, marking an 8% increase from the 882 outages reported in May. Similarly, the United States experienced an increase in outages, rising from 287 in May to 308 in June, a 7% increase.

Bar chart showing global and U.S. network outage trends over the past eight weeks.
Figure 6. Global and U.S. network outage trends over eight recent weeks

Subscribe to the ThousandEyes Blog

Stay connected with blog updates and outage reports delivered while they're still fresh.

Upgrade your browser to view our website properly.

Please download the latest version of Chrome, Firefox or Microsoft Edge.

More detail