% The Scalable Commutativity Rule

# administrivia

sign up done? see schedule

next: James W. & Naveen & Omar

paper questions/comments due 11pm before class

project proposal due 11pm next Friday

# paper discussions

notes: background information

end-to-end scenarios

# "example" paper questions

"where the hell is Farmington, Pennsylvania"

"Does having > professors:students lead to a qualitatively different paper?"

# scalability vs. performance

"A service is said to be scalable if when we increase the resources
in a system, it results in _increased performance_ in a manner
proportional to resources added."

-- Werner Vogels,
["A Word on Scalability"](http://www.allthingsdistributed.com/2006/03/a_word_on_scalability.html)

. . .

what's "scalability" in today's paper?

# multicore research overview

2006 [The landscape of parallel computing research: A view from Berkeley](http://www.eecs.berkeley.edu/Pubs/TechRpts/2006/EECS-2006-183.html)

2012 [The multicore evolution and operating systems](https://apsys2012.kaist.ac.kr/media/slides/apsys2012-keynote.pdf)

# kernel scalability

application scalability depends on kernel scalability

limiting factors: shared data, shared hardware

Corey (OSDI 2008): push to applications

Barrelfish (SOSP 2009): turn into message passing

Linux kernel analysis (OSDI 2010, OLS 2012):
fine-grained locking/
per-core partitioning/
reduce sharing/
scalable ds

examples:
affinity-accept (EuroSys 2012),
bonsai tree (ASPLOS 2012 - next Monday),
radix tree (EuroSys 2013)

interface design: today's paper (SOSP 2013)

see also:
[fos](http://groups.csail.mit.edu/carbon/?page_id=39),
[Tessellation](http://tessellation.cs.berkeley.edu/)