Massive Power Outage At Rackspace'S Dallas Data Center
Summary: The Rackspace data center in Dallas (DFW), where our servers are located, suffered a major power outage tonight lasting approximately two hours. A truck hit one of their transformers, causing damage, which compromised …
Context & Ripple Effects
On the night of November 12, 2007, a truck strike on a transformer took out power at Rackspace's Dallas (DFW) data center for roughly two hours — a confirmed physical failure, not a software fault. What makes the story travel is who was inside the building: the outage knocked out 37signals' own applications, prompting a downtime explanation on the company's product blog within hours, while GigaOM and Valleywag picked up the incident the same night.
Valleywag's framing — a Texas truck driver 'killing all the websites you really use' — captures why this matters beyond Rackspace's customer list: by 2007 a meaningful share of the recognizable web runs through a handful of managed-hosting facilities, so one transformer becomes a shared point of failure for sites with no operational relationship to each other.
First-order effects
- Rackspace's Dallas customers, including 37signals' project-management and collaboration apps, were unreachable for approximately two hours, forcing each dependent business to publish its own explanation to users.
- A single vehicle accident defeating the facility's power chain puts Rackspace's managed-hosting reliability pitch under direct scrutiny from exactly the high-profile customers it courts.
Second-order effects
- Rival managed hosts gain a sales opening around geographic and electrical redundancy, and multi-site customers will press Rackspace on whether DFW-style single-transformer exposure exists elsewhere in its footprint.
- Hosted-application vendors like 37signals face pressure to spread production across facilities or regions, since their brand absorbs the downtime even when the failure is their provider's.
Third-order effects
- If web infrastructure keeps consolidating into a few large data centers, mundane physical events — vehicles, weather, grid faults — become internet-visible outages, making facility-level redundancy a purchasing criterion rather than a datasheet line.
- The incident foreshadows the pattern behind later grid-interaction debates over data centers, where the boundary between a hosting company's power architecture and the public grid becomes an industry-level concern.
The trend: As hosting concentrates into fewer large facilities, single-site physical failures increasingly register as internet-wide outages, pushing redundancy decisions up from the server rack to the facility and region.