Delivering Applications Beyond the Edge

"The New Frontier is here, whether we seek it or not” – JFK, 1960

Since the very early days of the Internet, application load time has been a critical metric for all online businesses, who today spend several billions of dollars annually on different application delivery solutions to make sure their users are able to access their content online as quickly as possible. However, in spite of widespread awareness of the importance of speed, and concerted efforts by these businesses and vendors worldwide, the web in the last few years has slowed down, even though the Internet itself has gotten faster. This is a confounding paradox, until one takes a look at the underlying discontinuous shift in the application delivery problem and the assumptions that legacy solutions make.

The first generation of web users connected to the Internet from a “Wintel” box on a wired last mile. There were huge delays in the time it took for application content to reach the edge from the origin servers where the content was hosted. Seeing this as the key problem, several companies sprang up with CDN solutions, which addressed those delays by building out a system of internetworked caching servers that they placed at the edge. Users of companies deploying such a solution could then quickly grab content from the edge server closest to them, eliminating the latency bottlenecks in the core of the Internet. Since then, all application delivery solutions have chased building the biggest and fastest edge network, optimizing their system of caches, networks and storage for the last drop of cache lookup performance, shaving milliseconds in network roundtrips and leveraging the latest innovations in storage.

In a frantic race to optimize the edge network, the means ended up becoming an end unto themselves. CDN solutions simply ignored the fact that users have now stopped connecting to the edge directly but use wireless connections instead — which creates even more delay between them and the edge than there is between the origin server and the edge. The challenge now is not only how do we optimize application delivery to the edge, but also over these wireless last mile connections and onto the myriad devices users consume their applications on. In pursuit of increasingly rich and responsive mobile-first applications, users are fast moving away from the edge and are not coming back.

For the first time we are required to shift our perspective and look beyond the edge. The overall delivery path we need to optimize involves a system not just of the cache servers and transit network, but also last mile connections and application endpoints. We on the infrastructure side must tackle this problem end-to-end. The solution must work seamlessly across different kinds of physical clouds, networks and devices.

Instart Logic solves this problem by building an abstracted view of this entire system, taking a software-centric approach to the solution — where the features and optimizations are divorced from the underlying system configuration — and focusing on the application itself whose delivery needs to be optimized.

Today we are proud to fully share our vision for the future of application delivery solutions. Ever since launching our streaming features over the last year, our team had been working hard on features that extend beyond addressing the last mile congestion issues, and address the diversity challenge posed by combinatorial explosion of screen sizes, cache types and processing capability of different user devices. The ability to resize images dynamically for different screen sizes, transcoding images to the right format for best experience, and dynamically picking the right cache inside the client to optimize load performance are all instances of these.

Cloud applications are now increasingly the lifeline of today’s world, and their performance is critical to the economy. Over time the underlying networking and hardware technologies will improve as standards continue to evolve, but the growth in complexity of applications and increase in user expectations are going to far outstrip the pace of this improvement. An intelligent software-defined solution to application delivery is our best bet to stay ahead, and this is precisely what Instart Logic hopes to provide to the market.