USGS meets spikes in demand for earthquake info with content network
Content delivery network helps the U.S. Geological Survey keep its sites available even when traffic surges
The magnitude 9.0 Tohoku earthquake that took place March 11 near the northeast coast of Honshu, Japan, resulted in 6,000 hits per second on the U.S. Geological Survey’s earthquake-related websites, which are delivered via a content delivery network (CDN) service.
The peak Web traffic record achieved by USGS still stands at 52,000 hits per second following the 7.2 magnitude earthquake that occurred on Easter Sunday 2010. USGS now has additional metric data available on site traffic, which shows that earthquake sites received 400,000 visits and 1.5 million page views on an average day for the month surrounding the latest Honshu earthquake. The peak following the recent Japan earthquake and tsunami was 900,000 visits and 2.5 million page views per day.
According to Lorna Schmidt, program manager for the USGS Enterprise Web Program, earthquakes that occur in the United States, particularly in densely populated areas result in significantly larger spikes on our websites than those that occur outside the United States. Such massive spikes in data and information requests, produced by flash crowds who felt the earthquake, generally occur very soon after an earthquake event. “Such spikes, or peaks, generally have a precipitous drop off to a more smooth, elevated request rate for some period surrounding the event. This may be due to news coverage, family interest or other means of distributing information in the event,” she said.
USGS relies on a Level 3-supplied CDN to manage the flash crowds for earthquake data requests and provides cached access to data from globally distributed systems. Access to source data is critical to USGS’ primary mission, providing updated, refreshed content to citizens as well as internal programs and processes. The CDN allows USGS to get the job done with an infrastructure that is leased through a CDN contract, rather than owned, to reduce costs.
During the recent 9.0M Japan earthquake and tsunami, the Level 3 CDN performed bursting, which replicates requested data through the global network, serving increasingly more bandwidth to users until peak demand was satisfied. Then the CDN eased off replicating as demand drops back to more typical levels. USGS pays service fees both for DNS and bursting charges.
Data coming into USGS, such as earthquake sensor and streamflow data, travels multiple paths from thousands of stations to redundant systems to ensure delivery to the USGS emergency notification system and to the public. USGS Web content hosted within the NatWeb infrastructure, such as local science centers and many National USGS Program sites, is managed in a cloud of file and Web servers at three data centers across the United States. Each center’s servers are configured identically for backup purposes, with data replicated among all servers.
The current environment supports the agency’s continuity-of-operations plan and allows USGS to continue to provide access to natural resources data, no matter how great the demand. A CDN with a worldwide presence reduces geographic latency, or the distance between the requester and server, USGS officials explained.
Over the years, USGS has learned that testing the system is important. In addition, monitoring Web sites before, during and after an event can help improve resiliency. In all phases of the information life cycle, organizations must build and account for redundancy. USGS officials recommend that organizations design any Web infrastructure with performance demands and outages in mind and plan for the necessary redundancy upfront. The current server infrastructure, along with the use of a CDN, allows USGS to gather data and keep it available even when demand peaks during emergency situations.
As part of life cycle planning, the Enterprise Web Program and USGS will continue to evaluate strategies for delivery of USGS content and information, taking into account a fluctuating budget, security, performance and technology requirements.