F5 Synthesis: Avoiding network taxes to improve app performance

#SDAS #webperf Consumers are increasingly performance-sensitive, which means every microsecond counts.

In the five seconds it takes you to read this, 60% of your visitors just abandoned your site. Another 20% were gone before you even hit the first comma, and 30% of all them are purchasing from one of your competitors. That's what Limelight Networks "State of the User Experience" says, which pretty much falls in line with every other survey of consumers. They are, on the whole, impatient and unwilling to suffer poor performance when there is a veritable cornucopia of choices available out there on the Internet.

In an application world, apps equal opportunity. But that opportunity is the double-edge sword of Damocles, just as able to offer opportunity to your competitors as it is you.

Application performance has always been critical. So critical, in fact, that entire markets have cropped up with solutions to address the application and network problems that cause poor performance.

Ironically, many of them today may contribute as much to the problem as they do resolve it.

Most web performance optimization (WPO) or front-end optimization (FEO) solutions are pain point products. That means that when a pain point - like web application performance - becomes problematic enough for the business to notice, a solution is quickly acquired and deployed.

And thus begins the conga line of products designed to improve application delivery.

Each addition to the line introduces delay between the consumer and the application. Not because the products are slow or perform poorly, but because there are absolute minimums in terms of the time it takes to open and close TCP sockets and transmit packets over a wire. You can't eliminate it unless you eliminate the solution.

Such an architecture might not be problematic if it was just one or two services, but enterprises typically end up with a significant number line of devices in the critical data path between the consumer and the application.

Figure 1: Traditional conga-line deployment of application services

Each of these incurs a certain amount of latency necessarily due to network and protocol requirements and contributes to the responsiveness (or lack thereof) of applications. It's a network and protocol tax on each and every service that can't be avoided in this architectural model.

This is where a service platform can help. In addition to its strategic advantages, the platform approach to delivering application services through F5 Synthesis has positive performance implications, as well. When one platform can support the services you need to improve performance and enhance security, you can eliminate all the latency incurred by deploying those same services on multiple, disparate systems.

Figure 2: Modern, platform deployment of application services.

By building application services atop a common, high performance service platform, F5 can deliver a broad set of application services without incurring the network and connection taxes required. This eliminates many microseconds (or more, depending on the network topology and conditions) from the consumer response time, meaning applications are delivered faster and consumers don't ditch you for someone else.

A platform approach to application services has many advantages. Eliminating the network and protocol taxes required by traditional service deployment models is one them.

F5 Synthesis High Performance Service Fabric is built atop such a platform, and with the broadest set of application services available on a single platform, you can turn your application service infrastructure into a competitive advantage.

For more information on F5 Synthesis:

Published Aug 08, 2014

Version 1.0