Dec 1, 2014
While consumers were busy shopping on Black Friday, we were busy documenting their impact on the cloud. Teaming up with Internetweathermap.com we were able to capture end-to-end latency for hundreds of thousands of routes during the day from the across the Internet. We supplemented that research with packet loss statistics for major US backbones from Keynote.
Latency variations are particularly significant as latency is essential to the throughput of a TCP session. The higher the latency, the lower the throughput. As enterprises move large amounts of data to the cloud, latency increases will cause transfers t0 take significantly longer or fail completely. And while SaaS providers generally do an excellent job managing their own facilities, they have no control over the Internet provider’s network.
Routing policies and peering arrangements may direct traffic in the way that makes best economic sense for the provider, which may not necessarily — and often does not — deliver mean optimum application performance for the enterprise. These fluctuations are particularly significant in times of network congestion, such as on Black Friday. And with IBM reporting that online sales jumped by 9.5 percent year-over-year, SaaS providers saw dramatic increases in the time needed to reach their services within North America and globally.
Packets may be sent around the globe when they only need to travel across the city. Don’t believe me? How else do you explain latency on a route from New York, NY to Ashburn, VA where ServiceNow.com has its website, increasing 272 percent from 39 ms to 141 ms. ServiceNow automates and manages IT service relationships across the global enterprise. To put that in perspective, the additional 102 milliseconds is about the time needed to go between New York and California, let alone to Virginia.
Even within cities some routes were impacted. Microsoft’s Office.com saw latency increase by as much as 317 percent within Redmond, Washington. SalesForce.com didn’t escape the impact of Black Friday either with latency between Dallas and San Francisco jumping 358 percent, increasing from 97 milliseconds to 444 milliseconds.
With Black Friday’s impact growing overseas, international routes were also affected by the holiday. Delay changed the most between London and Denver, CO for Marketo, a leader in digital marketing software and solutions, increasing from 97 ms to 444 ms, a jump of 252 percent. The worst international region? South Africa, the origin for about a third of the most erratic 50 routes.
While few days may show as much traffic as Black Friday or Cyber Monday, fluctuations and latency remains a factor in any unmanaged network, such as the Internet. If organizations are to rely on the Internet to deliver predictable, critical application, IT leaders must plan for Internet Weather.
For more information about Internet weather during Black Friday and Cyber Monday follow #HolidayInternetWeather on Twitter or visit the Holiday Internet Weather blog. Full details about our methodology can be found here.