Sitting at a desktop, the average bank customer checks his or her accounts three or four times a month. A mobile customer, while waiting for a train, riding the train, sitting at a café or even waiting to pick up a child from soccer practice, will check his or her bank account 15 to 20 times per month.
This, according to Archie Roboostoff, results in more than a quadrupling of traffic to the site, stressing the infrastructure. Roboostoff, the Borland solutions portfolio director at Micro Focus, made the point to illustrate that mobile devices—and mobile access to websites and applications—are changing the expectations of back-end software.
It’s this “mobile multiplier” that Roboostoff believes is driving a change in the traditional three-tier infrastructure that defined applications almost since the inception of the Web. “The three-tier infrastructure—Web hed, app server, database—was great. You’d initiate a request and get a response. The response branched off into two-things: real-time acquisition of data, and then caching of things that I’m looking at over and over again. That was doing well for quite some time. We saw a number of customers try to leverage that same infrastructure for when they were putting out their mobile app. They tried to make it work with Web services that were wrapped, but one of the dynamics changing the need to do that is the notion of what we call the mobile multiplier.
“Let’s say, for simplicity, you’ve got 1,000 end users hitting your website on a daily basis,” he continued. “Now those end users typically have more than one device, and they’ll start interacting with your back end a little bit more often, especially if you’re getting into a customer intimacy application, where you’re wanting to push data more often. So the volume of data is increasing, the level of frequency of that data is increasing, and the problem is the infrastructure is set up for the requester of that data coming in on a very high-speed network, when in fact the majority of the data is coming in from a vastly slower network, a 3G or 4G network, which are still fast but not fast compared to a fiber-optic network. A lot of those issues…the need to have very quick responses for in-app purchases, the fact that security is still kind of an issue with REST-based services, that’s kind of driving the need to change how people think about how they implement their back-end infrastructure, [and] the technology they’re using to deliver that back-end infrastructure.”
Roboostoff pointed to such things as node.js, which creates smaller, localized servers, and Kaazing’s WebSocket Gateway, which is designed to fire off rapid requests over and over again—“almost like mini HTTP servers”—as ways to address these issues.
“We’re seeing customers setting up dedicated infrastructure for mobile traffic,” said Roboostoff. “Microsoft and IBM will make the necessary adjustments in their technologies eventually” to accommodate mobile applications, he said, and he’s seeing companies like AppFog, which offers a Platform-as-a-Service, making a number of acquisitions for higher-speed, higher-scale node.js deployments.
Another key technology Roboostoff cited is HTML5, which enables developers to give customers the look and feel that they’ve become accustomed to with their consumer applications.
For enterprise organizations, this means making sure their transaction gateways are strong enough to satisfy such things as in-app purchases; when players buy 100 coins in the middle of a game, you need to verify the coins are delivered without hanging up the game, he said.
Further, these organizations must be able to test their applications across every platform, not only functional and load testing, but also doing network simulation to ensure the application will perform optimally regardless of where the request is coming from.
“We’re seeing a lot of customers not only rethink how they’re doing their infrastructure planning, but also their pre-production load testing,” he said. “Traditionally, you put out a Web server, send a sustained user load, see what happens, and fine. But customers are asking us to implement being able to simulate network-loading patterns. They want 100 users to come in at 9 a.m. on a 4G network, followed by everyone on a T1, then after lunch have users coming in on a BlackBerry, to see what’s happening and why. Customers are still in learning mode to see how they’re doing their infrastructure on the back end to accommodate those traffic patterns.
“Don’t just think you can throw a mobile app out there, hope for the best, and hope that the existing infrastructure works,” he said. “You’re going to need to think it through.”
David Rubinstein is editor-in-chief of SD Times.