No abstract

The past few years have seen a dramatic increase in the Internet traffic due to popular bulk data transfer applications, such as BitTorrent. Many access ISPs are deploying traffic shaping to limit the peak network traffic on their transit links, and thereby reduce their wide-area bandwidth costs. In this paper, we show that an ISP can substantially reduce (by a factor of 2 or more) its peak transit link usage by traffic shaping a small fraction of its largest flows, while incurring a minimal penalty on the completion times of these bulk flows. Unfortunately, our analysis also shows that if many ISPs traffic shape flows based on their local interest, the end-to-end performance of the bulk flows will be seriously harmed. Categories and MOTIVATIONThe Internet is witnessing explosive growth in demand for bulk content. Examples of bulk data transfers include downloads of music and movie files and distribution of large software amongst others. Recent studies of Internet traffic [1,8] show that such bulk transfers account for a large and rapidly growing fraction of bytes transferred across the Internet.The bulk data traffic in the Internet today represents just the tip of the iceberg. Tremendous amounts of digital data are being delivered outside the Internet, for example using hard drives, optical media, or tapes [5,6,9], because it is cheaper and faster -though usually not more convenient or secure -than using the Internet. For example, on an average day, Netflix ships 1.6 million movie DVDs [9], or 6 petabytes of data. This is more than the estimated traffic exchanged between ISPs in the USA [10]. It is debatable whether the Internet can ever match the capacity of postal networks. However, the convenience of online transfers is likely to drive the demand to deliver more bulk data over the Internet in the foreseeable future.Internet bulk data transfers are expensive. A recent study [7] reported that ISPs (or CDNs) charge large content providers, such as YouTube and MSN Live, 0.1 to 1.0 cent per minute for a 200-400 kbps data stream. And higher bandwidth streams will cost even more. The high cost of wide-area network traffic means that increasingly economic rather than physical constraints limit the performance of many Internet paths. That is, even when there is plenty of physical capacity available on a given link, ISP policies of charging customers based on peak bandwidth utilization (often measured by the 95th percentile over some time period) result in strong disincentives to approach the full physical capacity of inter-ISP links.While decades of research in congestion control shows how to manage transfers across physical bottlenecks, there is little understanding of how to manage transfers across economic bottlenecks. Instead, ISPs have developed a variety of ad hoc traffic shaping techniques to control bandwidth costs. This traffic shaping typically targets bulk transfers as they consume the majority of network bytes. However, the techniques are often blunt and arbitrary, and often shut down entire appl...

No abstract

Realizing adaptive and efficient peer-to-peer Service-oriented Architectures for MANET environments is a challenging problem. In particular, robust and efficient service discovery and service migration are critical in the constantly changing and bandwidth limited MANET environments. In these scenarios, service migration lays the foundation for self-adaptive architectures. This paper describes the agile computing approach to peer-to-peer service discovery and service migration and provides a performance evaluation of these functions in the context of the Agile Computing middleware. The experimental results presented in the paper show that applications built on top of the Agile Computing middleware are capable of opportunistically exploiting transient computing resources in the MANET environment.

The Internet is witnessing explosive growth in traffic due to bulk content transfers, such as multimedia and software downloads, and online sharing of personal, commercial, and scientific data. Yet bulk data transfers remain very expensive and inefficient. As a result, huge amounts of digital data continue to be delivered outside of the Internet using hard drives, optical media or tapes. Meanwhile, large reserves of spare bandwidth lie unutilized in today's networks, where links are overprovisioned for peak load. We designed NetEx, a bulk transfer system that opportunistically exploits the excess capacities of network links to deliver bulk content cheaply and efficiently. Our results based on data from both a commercial tier-1 ISP and the Abilene network suggest that NetEx can considerably increase the capacity of the network, and at the same time it can provide good average performance to bulk transfers.