The Local and Global Effects of Traffic Shaping
نویسندگان
چکیده
The Internet is witnessing explosive growth in traffic due to bulk content transfers, such as multimedia and software downloads, and online sharing and backups of personal, commercial, and scientific data. Wide-area network bandwidth is expensive and this cost is forcing many ISPs to deploy middle boxes to contain bulk traffic. As a result, many Internet bottlenecks today are economic rather than physical. That is, for many links interconnecting distinct organizations, there is often plenty of available physical capacity. However, the cost of actually transmitting across these link is based on peak levels of utilization, for instance as measured by the 95% link utilization. Thus, there are incentives to perform traffic shaping across these links to limit peak levels of utilization. In light of these trends, this paper makes the following contributions. We show that appropriate inter-ISP traffic shaping mechanisms can dramatically reduce peak levels of utilization with no impact on interactive applications and only minimal degradation of bulk data transfers. This suggests that in the future, it is in the self-interest of every ISP to perform such traffic shaping at the edges of its network. Unfortunately, we show that the local incentives to perform traffic shaping will result in dramatically negative global slowdown of bulk transfers, with the degradation growing as a function of the distance between the source and destination. Essentially, geographic time zone differences and the associated offsets in the local times of peak utilization mean that bulk transfers traveling sufficiently far will always be throttled by some ISP between the source and destination. Our findings suggest that once traffic shaping becomes predominant, alternative data transfer mechanisms will be needed to efficiently deliver bulk data across the Internet.
منابع مشابه
Effect of social traffic behavior on traffic violation the case study of Tehran
Current standards in microscopic modeling lack the ability to address certain traffic scenarios, the ability to capture discrete differences in driver behavior that exist between various driving populations, and are structured in such a fashion that various vehicle movements are predicted and modeled separately. In this study, effect of social traffic behavior on rate of traffic violation of ...
متن کاملLocal and Global QoS-aware Token Bucket Parameters Determination for Traffic Conditioning in 3rd Generation Wireless Networks
Based on the framework of employing traffic shaping at the UE and policing at the RNC, we propose a local and global QoS-aware token bucket parameter determination technique for traffic conditioning in 3G wireless networks in this paper. The local QoS-awareness is achieved by the tradeoff between the shaping delay and the out-ofprofile probability at the token bucket traffic shaper, and the glo...
متن کاملThe Local and Global Effects of Traffic Shaping ( Extended Version )
The Internet is witnessing explosive growth in traffic, in large part due to bulk transfers. Delivering such traffic is expensive for ISPs because they pay other ISPs based on peak utilization. To limit costs, many ISPs are deploying ad-hoc traffic shaping policies that specifically target bulk flows. However, there is relatively little understanding today about the effectiveness of different s...
متن کاملAn investigation on the social and traffic effects of metro stations on neighborhoods
Neighborhoods within the hierarchy of urban structure are one of the smallest socio-physical building blocks of cities. Traditional neighborhoods of Iranian historic cities benefited from strong social relations which were flourished over the years and were manifested in the neighborhoods centers. In the modern era, the new master plans and the complementary local plans have disintegrated th...
متن کاملMultiagent Learning with a Noisy Global Reward Signal
Scaling multiagent reinforcement learning to domains with many agents is a complex problem. In particular, multiagent credit assignment becomes a key issue as the system size increases. Some multiagent systems suffer from a global reward signal that is very noisy or difficult to analyze. This makes deriving a learnable local reward signal very difficult. Difference rewards (a particular instanc...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2009