Internet paths / end systems / web portals / ratio contrasting end-system energy consumption / intensive applications / network energy / per-request energy consumption / energy impact / high baseline energy consumption / per-request energy consumptions / greater energy efficiency / p2p applications / work-induced energy / average Internet path length to/from / Energy-proportional machines / per-bit energy increase / p2p systems / energy / energy costs / per-bit router baseline energy / per-bit baseline energy consumption / Internet applications / network/routers / Internet traffic / energy consumption / energy components / baseline energy consumption / per-bit server baseline energy / end system energy / maximum web server capacity / p2p protocol / energy scales / decentralized systems / non-network energy component / energy equations / web search / energy efficiency / energy-efficient data-center design / per-bit energy / networked systems / air-conditioning systems / work-induced energy consumption / baseline energy / /
Hot Data Centers vs. Cool Peers Sergiu Nedevschi† , Sylvia Ratnasamy† and Jitu Padhye‡ † Intel Reserch, ‡ Microsoft Research 1 Introduction Most networked applications use either a largely centralized architect