3rd Party Remote Call Monitoring Feature

A recent report conducted by the European Union Agency for Network and Information Security (ENISA) found that two-fifths of mobile Internet outages happened because of software bugs—a significant increase over the 2012 total of just 15 percent. Anecdotal evidence suggests this may also be a problem in the U.S.

According to Gartner (News - Alert) research director Sylvain Fabre, mobile networks depend more on software than before but that software is not getting tested as much as it should. This results in outages that can impact millions of connections.

This dependence is out of necessity as software defined networking (SDN) provides a cost effective way for mobile network providers to manage upgrades and is highly scalable. When a network grows past a certain level, a hardware-intensive architecture becomes costly and impractical to upgrade after new standards and protocols are released. An SDN architecture that virtualizes the functionality can be updated and scaled much faster.

The demand for SDN is not going away anytime soon either. International Data Corporation (IDC (News - Alert)) predicts an eight-fold increase in the global SDN market for cloud provider and enterprise segments, growing from $960 million this year to over $8 billion in 2018.

An outage that affected Facebook’s (News - Alert) site back in June was caused during a configuration update of a software system. Less than a month ago, Time Warner suffered a nationwide outage that, according to a Popular Science interview with Purdue University (News - Alert) computer scientist Sonia Fahmy, was likely caused by a mistake while configuring software for routers. Although neither case deals with mobile networks, software was either a probable or definite cause.

Software that does anything meaningful is complex by nature and trying to make it perfect is a Sisyphean task at best. You cannot test enough because not every possible outcome can be tested. Once you find a bug, fixing it could introduce more bugs and perpetuate the cycle, but if Fabre’s conclusions about testing are true globally, that’s a problem. Software that affects networking that millions of users rely on should be tested thoroughly before it is released into production.