Table 1: Effect of Distance on Throughput and Download Time Distance (Server to User) Network RTT Typical Packet Loss Throughput 4GB DVD Download Time Local: <100 mi. 1.6 ms 0.6% 44 Mbps (high quality HDTV) 12 min. Regional: 500–1,000 mi. 16 ms 0.7% 4 Mbps (basic HDTV) 2.2 hrs. Cross-continent: 3,000 mi. 48 ms 1.0% 1 Mbps (SD TV) 8.2 hrs. Multi-continent: 6,000 mi. 96 ms 1.4% 0.4 Mbps (poor) 20 hrs Although many alternate protocols and performance enhancements to TCP have been proposed in the literature ([23], [30], [45]), these tend to be very slow to make their way into use by real-world end users, as achieving common implementation across the Internet is a formidable task. Scalability. Scaling Internet applications means having enough resources available to respond to instantaneous demand, whether during planned events or unexpected periods of peak traffic. Scaling and mirroring origin infrastructure is costly and time-consuming, and it is difficult to predict capacity needs in advance. Unfortunately, underprovisioning means potentially losing business while overprovisioning means wasting money on unused infrastructure. Moreover, website demand is often very spiky, meaning that companies traditionally needed to provision for anomalous peaks like promotions, events, and attacks, investing insignificant infrastructure that sits underutilized most of the time. This also has an environmental cost when underutilized infrastructure consumes significant amounts of power [33]. Finally, it is important to note that origin scalability is only apart of the scalability challenge. End-to-end application scalability means not only ensuring that there is adequate origin server capacity, but also adequate network bandwidth available at all points between end users and the applications they are trying to access. As we will discuss further in Section 5.1, this is a serious problem as Internet video comes of age. 100>