iOS 7 Update: A Lesson In Event-Driven Network Spikes update from September 2013

Network managers around the world undoubtedly knew something was up around 10 a.m. Pacific Time Sept. 18, a date Apple users circled on their calendars. Indeed, it took all of 38 minutes for the North American Network Operators Group (NANOG) mailing list to light up with messages about traffic spikes, undoubtedly caused by millions of users updating to iOS 7. And yes, it was millions of users, given the installed base of over 600 million iOS devices with a sizeable share of those in North America.

While it's easy to point the finger at Apple or its impatient and ignorant (at least of network limitations) users, this type of event should be a wake-up call for network managers. It's a predictable consequence of the rapid, inexorable shift of software and content distribution from physical media or broadcast signals to the cloud. Whether it's binge watching "House of Cards," streaming the news during a major incident like the Boston Marathon bombing or checking out a live feed from the Olympics, the age of event-driven network spikes is upon us.

Although the iOS upgrade had strength in numbers, it was a relatively tiny payload--750 Mbytes (iPhone) to 950 Mbytes (iPad)--in comparison to HD video streams that Netflix says use about 2.8 Gbytes per hour, which equates to about 6.3 Mbps per stream. Bottom line: The problem will only get worse.

Online ad network Chitika found that a mere 48 hours after the iOS 7 release, nearly 32% of North American iOS Web traffic came from iOS 7 devices. Mixpanel, a mobile app analytics firm, found that iOS 7 was on 22% of iOS devices in the first 10 hours, meaning many people did upgrade from work. So there were undoubtedly tens of millions of iOS upgrades happening during the workday. That's a lot extra network traffic, which meant network managers everywhere were seeing traffic spikes similar to this one posted by a NANOG member.

Of course, the Internet backbone didn't feel a thing since most of the traffic came from local Akamai caches staged at ISPs everywhere. The content delivery network (CDN) did see some hot spots, particularly in Europe. Numerous NANOG members reported large traffic spikes, with one participant writing, "Our local Akamai cluster has pegged its 1G uplink a few times, and we are hitting our 1G Equinix IX [Internet Exchange] link pretty hard as well." But in general, their local CDN caches and large inter-exchange pipes weathered the storm.

[Video traffic also is putting pressure on networks. Read how IT teams need to prepare for the onslaught in "Video Evolution: Brace Yourself For Impact."]

Campus and enterprise networks didn't always fare so well. According to Ohio University's student newspaper, "an outage across campuses kept students and faculty from logging on. The Internet went down at about the same time Apple’s new mobile software, iOS 7, became available in the area." A similar outage hit the University of Texas Pan American campus. A report from its campus newspaper cited an IT email statement identifying the cause as "network saturation," with a follow-up message explicitly attributing the saturation to the iOS upgrade.

But instead of whining, network managers should use the iOS 7 incident as a call to action. If you're a service provider or large enterprise, one NANOG participant had sound advice: " Use it as an excuse to upgrade your pipes, talk Akamai or CDN of choice into putting a cache on your network, or implement your own caching solution. As operators of the Internet we should be looking for ways to enable things like this, not be up in arms at Apple for releasing an update to their phone OS or making it available in a way that's inconvenient to our oversubscription policies."

While a CDN isn't feasible at smaller organizations, local caching and QoS certainly is. While a cache might not have helped clogged edge networks, it does protect your WAN links. And with software appliances like Dell's SonicWall WXA 500 available for under $500, WAN optimization is feasible even at small branch offices.

Similarly, prioritizing traffic, something every enterprise router or UTM appliance can do, will ensure that mission-critical applications get what they need and maybe even force a few iPhone users into waiting until they get home before firing up the update process.

Apple updates iOS like clockwork every year. Use last week's update as a milestone to make sure your network doesn't miss a beat when the next one rolls around.

[Find out how to structure QoS in a Cisco environment that runs converged voice, video and data in Ethan Banks' session "How To Set Up Network QoS for Voice, Video & Data" at Interop New York Sept. 30-Oct. 4.]