Tutorial on latency at Sigcomm 2017
(too old to reply)
Joe Touch
2017-06-13 18:07:07 UTC
Hi, all,

I'll be presenting a tutorial on latency at Sigcomm 2017 in Los Angeles,
and invite you all to consider attending.

Please forward this notice as useful.

(COI Disclaimer: Sigcomm tutorial presenters are unpaid, FWIW)



ACM SIGCOMM 2017Tutorial
Understanding Latency - A Root Cost and Mitigation Approach
Joe Touch, USC/ISI
Call For Participation - Los Angeles, CA, Friday August 25, 2017

“Time is fleeting,” it has been said. Time delay, or latency, is the one
metric that drives most others. It defines what it means to be *fast*,
and limits us to what is *fast enough*.

Latency has always been a key part of network performance, but recently
it has been elevated to a primary focus for electronic traders, search
engines, name servers, data centers, and home and network routers.
Increases in network bandwidth, router forwarding speed, and end system
computational resources have helped bring latency to the forefront as a
primary concern. New protocols are emerging to address latency as a
primary issue, including Delay Tolerant Networking (DTN) and bufferbloat

This tutorial presents a comprehensive exploration of the impact of
latency on communication. It explores time as a budget to be spent,
over-spent, rebated, and conserved. We explore the root causes of
latency: generating data, transmitting it, processing it, and the impact
of resource sharing through multiplexing and aggregation, as well as
corresponding ways to mitigate each of these causes. We also explore
ways to mask latency that cannot be reduced and to avoid incurring its
cost in the first place. This tutorial provides examples of real system
design and implementations, as well as exploring several key case
studies to provide practical experience that attendees can immediately
apply to their own systems.