May 3, 2012

Why TCP acceleration matters to the CDN – Part 1


CDN for data and content deliveryI often get asked about how a Content Delivery Network (CDN) works and how companies can get the most use out of this service. In my three-part post, I will be answering those questions and explaining several ways to optimize the performance of a CDN. To begin, I’ll discuss what CDN and Transmission Control Protocol (TCP) acceleration technologies are and why optimization is important. In part two, I’ll explain how TCP acceleration applies to a CDN, and finally in part three I’ll take you through a use case that brings it all together.

What is a CDN?
A CDN is a cloud application built to deliver data faster to higher volumes of users across the world – whether it’s audio or video files, websites, game downloads, software patches or live streaming. Its goal is to accelerate content delivery and make it more scalable. This in turn contributes to a superior experience for end users and keeps them returning to your websites and online stores.

What is TCP Acceleration?
When TCP was introduced nearly 40 years ago, it was designed to facilitate reliable communication between two hosts. However, networks now operate at much higher speeds and regularly span much larger distances between the application origin and the end user, creating the need for a technology called TCP acceleration. Acceleration generally works at the TCP layer in one of two ways: symmetrical or asymmetrical. Symmetrical acceleration requires either two acceleration devices, one at each end of a network link, or a central acceleration device and client software that live at the end nodes. Asymmetrical acceleration has no dependency for a client device or software, rather it works from the server side alone. Typically, symmetrical acceleration performs better than asymmetrical, since it can compress the content in ways that cannot be done otherwise.

Why is optimization important?
While CDNs inherently accelerate content, the explosion of bandwidth-intensive data traveling over the Internet and the more global distribution of customers, is exposing the need to increase speed and address latency challenges within specific points on a CDN network. In fact, Cisco predicts by 2015, one million minutes of video content will traverse the Internet every second.* In order to prepare for this surge in content, every component of a CDN must be optimized − from uploading content to the final delivered “bit” of a game download. The elements that make up a CDN must perform in harmony to create the lowest-latency experience possible.

One aspect of the CDN that is often overlooked is the TCP layer of the network. CDNs, for the most part, move all their bits across TCP/IP, and both HTTP and RTMP traffic moves across this transport layer. By accelerating this layer to perform at its peak potential, fundamental design limitations of TCP that regulate and diminish web traffic performance can be overcome, allowing the CDN to achieve heightened performance.

Whew, a lot to digest huh? While you let that marinate, skim our State of CDN and WTH is Up With Your WPO infographics, which take you through the 1-2-3 of CDN and other website performance optimization technologies. Then check back here tomorrow as I detail how TCP acceleration applies to the CDN.

*, “Cisco, RealNetworks, Adobe, Harmonic: Adaptive Streaming and HTTP,” July 26, 2011.

Explore HorizonIQ
Bare Metal


About Author


Read More