How to make selfish apps share

There's is an insidious threat creeping into networks that goes largely unnoticed. One that is leeching the life out of your corporate network. Are you aware of what hungry, 'selfish applications' might be inconsiderately chewing through your bandwidth? 

So what is a selfish application? Its one that tends to utilise all available network bandwidth when being downloaded for the first time or when performing a software update. For example, if the download is coming over a 3 Megabits per second WAN link to a branch office, a Selfish Application would take all of it for the duration of the download. If it's a 20 Megabits per second Internet gateway, all of the bandwidth could be consumed for a download, although for a much shorter period of time. Obviously, this has the potential to seriously disrupt business operations if not managed properly.

When downloading begins, bandwidth consumption spikes upward without warning and creates contention with existing application traffic, drowning out other programs from using the network. The impact on real time applications that are especially sensitive to latency and contention is heavy – things like video conferencing, voice-over-IP and virtual desktop (VDI) sessions. Other non-real-time applications also suffer. Enterprise transactions slow or stop. Oracle or SAP response times become terrible.

There are actually two main problems that come from this. The first, and most severe, is the impact on application performance over the network. The second is the unpredictability of the problem.

This is a fairly common issue and typical of BYOD software associated with tablets and handhelds as well as almost any consumerised IT application. It is also the case with most utility software, such as Adobe Flash or Adobe Acrobat updates. Almost every company has felt the effects of selfish applications, but lacked the visibility or understanding of what the cause might be.

The fact is that more and more applications are becoming selfish and without a doubt, the problem is becoming worse. The increasing trend toward BYOD brings more mobile devices onto the corporate network, and all of the initial software downloads, the addition of new apps and frequent cycle of updates will greatly expand the effect of selfish applications. In addition, accelerated adoption of consumerised IT applications will extend the effect. Furthermore, the tendency for increasingly frequent, and bloated in size, updates will compound the problem greatly.

It's hard enough to develop a stable application these days that provides great utility and usability, security and reliability without having the network in mind. Many developers lack the time or resource to conduct real network testing.

At the same time being “selfish” is not really inherently bad. If you are the only user and that is the only application running, you would want the download to conduct itself quickly. The trouble is that on a network with lots of users and lots of applications, being selfish causes significant issues. The network itself really needs to intervene to optimise and manage application traffic. With the right technology, selfish applications can learn to play well with others. The selfish tendencies are needlessly selfish.

The first step is to have true application visibility to understand what is running on the network and how it behaves. The second step is optimisation to mitigate the effect of the downloads on the network. The third is to apply bandwidth management or Quality of Service (QoS) controls to prevent selfish behaviors taking over and draining the company network.

The key is to have a very granular view of traffic in terms of specific applications and then to be able to intelligently apply controls. Controls such as limits and priorities are important in managing selfish applications.

Another important solution is a next-generation WAN optimisation solution or Secure Web Gateway appliance that cache software updates locally. That means that a company's Internet gateway or WAN link to a branch office would only have to incur the cost of downloading an application or an update a single time. For instance, if there are 200 users in an office, and they all need an Adobe Flash or iTunes update in a short period of time, the network would only have to sustain a single download rather than repeating the same thing 200 times. In the case of an 80 megabyte file, reducing the “multiplicative” effect is practically revolutionary. In one case the total load is about 16 gigabytes; in the other it is only 80 megabytes.

Like small children, selfish applications are often innocently selfish. But with the right guidance in place, this selfishness can easily be negated.  

Greg Singh is the systems engineering manager of Blue Coat ANZ