UTCS FACULTY CANDIDATE: Kyle J. Nesbit/University of of Wisconsin - Madison Virtual Private Machines: A Resource Abstraction for Multicore Computer Systems ACES 2.302 Tuesday March 25 2008 11:00 a.m.

Contact Name: 
Jenna Whitney
Date: 
Mar 25, 2008 11:00am - 12:00pm

There is a sign up schedule for this event:
htt

p://www.cs.utexas.edu/department/webevent/utcs/events/cgi/list_events.cgi
Type of Talk: FACULTY CANDIDATE

Speaker/Affiliation: Kyle J.
Nesbit/University of Wisconsin - Madison

Date/Time: Tuesday March
25 2008 11:00 a.m.

Location: ACES 2.302

Host: Steve Keck

ler

Talk Title: Virtual Private Machines: A Resource Abstraction fo

r Multicore Computer Systems

Talk Abstract:
The computer industry
is undergoing a momentous
transformation. General-purpose computing is
moving
from desktops to diverse devices such as smart phones
digi

tal entertainment centers and data center servers.
At the same time h

igh-performance semiconductor
manufacturers have shifted their focus fr

om large
monolithic processor designs to distributed multicore
arch

itectures. In multicore architectures concurrently
executing threads o

ften share costly micro-architecture
resources (e.g. onchip storage an

d SDRAM memory
bandwidth) thus improving overall resource efficiency.

However contemporary operating system policies are
oblivious to mi

cro-architecture resource sharing and
contemporary multicore hardware

does not provide
the necessary mechanisms for operating systems
to

manage micro-architecture resource sharing.
Consequently contemporary

system architectures
are incapable of efficiently satisfying the divers

e
requirements of future systems.

In this talk I will present t

he Virtual Private Machine
(VPM) framework. The VPM abstraction provide

s the
conceptual interface between a multicore system%92s
resource

management policies and mechanisms. A
VPM consists of a complete set of
micro-architecture
resource assignments and a temporal processor
a

ssignment. VPM policies implemented primarily in
software translate t

asks%92 performance requirements
into VPM assignments. Then VPM mechani

sms
implemented in both hardware and software satisfy
the VPM ass

ignments. In effect VPMs virtualize a
multicore chip%92s performance a

nd provide tasks with
performance isolation. To illustrate the potentia

l of the
VPM framework I will present a set of VPM policies
and me

chanisms that seamlessly integrate micro-
architecture resource sharing

with common processor
scheduling algorithms.