BibTeX
@MISC{_reducingweb,
author = {},
title = {Reducing Web Latency: the Virtue of Gentle Aggression},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
To serve users quickly, Web service providers build infrastruc-ture closer to clients and use multi-stage transport connections. Although these changes reduce client-perceived round-trip times, TCP’s current mechanisms fundamentally limit latency improve-ments. We performed a measurement study of a large Web service provider and found that, while connections with no loss complete close to the ideal latency of one round-trip time, TCP’s timeout-driven recovery causes transfers with loss to take five times longer on average. In this paper, we present the design of novel loss recovery mech-anisms for TCP that judiciously use redundant transmissions to minimize timeout-driven recovery. Proactive, Reactive, and Cor-rective are three qualitatively-different, easily-deployable mecha-nisms that (1) proactively recover from losses, (2) recover from them as quickly as possible, and (3) reconstruct packets to mask loss. Crucially, the mechanisms are compatible both with mid-dleboxes and with TCP’s existing congestion control and loss re-covery. Our large-scale experiments on Google’s production net-work that serves billions of flows demonstrate a 23 % decrease in the mean and 47 % in 99th percentile latency over today’s TCP.
Keyphrases
web latency gentle aggression timeout-driven recovery large-scale experiment reconstruct packet latency improve-ments web service provider today tcp ideal latency google production net-work client-perceived round-trip time large web service provider congestion control easily-deployable mecha-nisms measurement study multi-stage transport connection novel loss recovery mech-anisms loss re-covery redundant transmission current mechanism round-trip time percentile latency infrastruc-ture closer
