In my last post, I got you up to speed on Content Delivery Networks (CDNs) and Transmission Control Protocol (TCP) acceleration. Now, I’ll take you through how TCP acceleration can be applied to three areas of a CDN network.
Ingest into the CDN
Getting content into the CDN can be a painful process if you have massive amounts of large content. Strategies like RYNC and FTP drop folders, which are used to manage the ingress of content, are common. Upload managers that utilize UDP and other forms of acceleration are also available for the purpose of getting very large files into the CDN.
For a website that uses a CDN, the ability to set up Origin Pull (Origin Pull is the method of transferring data to the CDN automatically from a web server as opposed to manually uploading the content) alleviates some of this pain; however, the interface for Origin Pull must be sophisticated enough to allow the operator to control the cache headers in order to make the site perform correctly.
Intra-CDN communication
Many CDNs incorporate acceleration between Points of Presence (POPs). This is often a form of WAN acceleration, which can take advantage of “both sides” of the communication chain. As a result, CDNs can be optimized even further. (Generally, symmetrical acceleration is better than asymmetrical acceleration since the former can compress the content in ways that cannot be done by the latter).
For example, replication of the content from North America to Europe or to the Pacific-Rim region is a perfect use case for this type of acceleration. Without acceleration, the files can take too long to get to their destination.
Egress acceleration
Egress acceleration is perhaps the most compelling area to apply acceleration within a CDN. Why? Because egress acceleration can have the most impact on a CDN’s users. Users surfing a website or downloading a multiplayer online game experience the advantage of the acceleration directly. For instance, TCP acceleration in this part of the CDN value chain has been found to accelerate the download of a large file over a high-latency connection by up to four times.
In real numbers that translates to downloading a 1MB file over a 200ms connection up to four seconds faster. While this may not seem significant, four seconds is an appreciable period of time on the Internet, and in most cases is more than enough time for a user to become impatient and close their browser window.
Be sure to stay tuned for my third and final post on CDN and why TCP acceleration matters next Tuesday, which will take you through an actual use case. In the meantime, check out our CDN Buyer’s Guide eBook
