|
Edge proxy caching networks—traditional content distribution networks (CDNs)—which were designed and optimized for caching small image and HTML files, have been highly effective over the last decade in pushing the Internet forward. Digital Island, Akamai, and Speedera were pioneers in this space. Without them, our experience on the early Internet would have been unbearable. The first generation
of CDNs improved Web page load times by storing content on servers closer to the end-user which eliminated the latency caused by the repetitive long round-trips that was required to fetch a Web page from the originating Web server. To make this effective and inexpensive across a large audience of online users, servers needed to be hosted on the ISPs network at each of the points where users dialed in or points of presence (PoPs). With thousands of ISPs, each with hundreds of POPs, tens of thousands of servers were required for a CDN to provide adequate coverage.
Given the scale required to maintain CDNs across multiple ISPs and the evolution
of the marketplace, including the growth of the Internet, increasing downloading
speeds, increased average file sizes, and newer applications such as video on demand, IP to television, and live broadcasting, Edge caching networks present significant drawbacks for companies and organizations that want to affordably provide fast-performing and reliable high-fidelity, interactive experiences
with streaming video, live broadcasting, and interactive applications.
This white paper outlines the limitations of Edge caching networks given the historical problem they solved and current trends, presents the attributes and benefits of an ideal solution, and suggests that BitGravity has designed a CDN with the attributes and benefits that provide companies and organizations with a superior solution that addresses the needs in marketplace at an affordable price.
Early Influences of the Internet on CDN Architecture
Prior to the existence of Edge caching networks, the Internet had to grapple with a myriad of problems that slowed the delivery of content over the Web.
› Limited Peering Points
Limited peering points between networks in North America were a culprit of poor performance. In the early to mid-1990s there were only two points where all the networks could connect their traffic, MAE-East and MAE-West. (MAE, an acronym for Metropolitan Area Ethernet, was an Ethernet set up in the Washington, DC area to peer and exchange traffic in the early 1990s. At the time, the MAE was the center of the Internet universe -- the one place where everyone came together and exchanged routes. After the success of MAE in the East coast, a similar facility was established in Silicon Valley, MAE-West.)
With only a couple of exchange points, data had to travel great distances
before getting to its ultimate destination. Could you imagine if California and Virginia were the only two points where all the freeways in the US connected? Something as simple as driving from Des Moines, IA to Wichita, KS would require driving to California first to switch onto the right freeway. Not the model for efficiency, but that’s what the Internet
looked like back then.
Hardware Latency
Dialup access was a big reason latency issues existed on the Internet ten years ago. Dialup access was not only limited in its bit transfer rates, but it also carried a high latency burden. The delay to transfer bits through the dialup infrastructure was sometimes as high as the latency for a bit traveling coast to coast at the speed of light across the wire.
› High Packet Loss
High packet loss led to poor performance and was symptomatic of the equipment at the time. With packets traversing the Internet long dis-tances, unreliable packet delivery usually meant retransmission of data or failure to deliver a file.
› Expensive Bandwidth Costs
Expensive bandwidth costs influenced behavior of the entire industry. The Internet was not yet mature enough to manage bursting traffic patterns. Companies compensated by over-provisioning and purchas-ing circuits with excess bandwidth to ensure unpredictable traffic could be handled. ISPs compensated by purchasing large over-provisioned circuits from the backbone carriers, and the backbone carriers over-provisioned to try to deliver content at reliable rates.
|
