Agenda

Monday December 14th 14:00-17:00 (room C669) ============================================

14:00-14:45: Measuring Serendipity: Connecting People, Locations and

Interests in a Mobile 3G Network

Speaker: Ionut Trestian (Northwestern University)

Paper: http://www.cs.northwestern.edu/~ict992/docs/serendipity.pdf

Characterizing the relationship that exists between people's

application interests and mobility properties is the core question

relevant for location-based services, in particular those that

facilitate serendipitous discovery of people, businesses and

objects. In this paper, we apply rule mining and spectral

clustering to study this relationship for a population of over

280,000 users of a 3G mobile network in a large metropolitan

area. Our analysis reveals that (i) People's movement patterns are

correlated with the applications they access, e.g., stationary

users and those who move more often and visit more locations tend

to access different applications. (ii) Location affects the

applications accessed by users, i.e., at certain locations, users

are more likely to evince interest in a particular class of

applications than others irrespective of the time of day. (iii)

Finally, the number of serendipitous meetings between users of

similar cyber interest is larger in regions with higher density of

hotspots. Our analysis demonstrates how cellular network providers

and location-based services can benefit from knowledge of the

inter-play between users and their locations and interests.

 14:45-15:30: Mitigating the Lack of Memory Safety
Speaker: Periklis Akritidis (Cambridge University)
C facilitates high performance and systems programming but lacks
memory safety, which undermines security and reliability: simple

bugs can breach security, and faults in kernel extensions can

bring down the entire operating system. These problems persist

because comprehensive solutions are either too slow for practical

use, or break backwards compatibility by requiring porting or ABI

changes. My talk will describe efficient backwards-compatible

solutions to mitigate the lack of memory safety. Improvements

over the state of the art were possible through judicious

trade-offs between performance and protection, careful

engineering, and focusing on enforcing only the minimum necessary

guarantees.

15:30-16:15: Sierra: a power-proportional, distributed storage system

Speaker: Eno Thereska (Microsoft Research)

I'll present the design, implementation, and evaluation of Sierra:

a power-proportional, distributed storage system. I/O workloads in

data centers show significant diurnal variation, with peak and

trough periods. Sierra powers down storage servers during the

troughs. The challenge is to ensure that data is available for

reads and writes at all times, including power-down

periods. Consistency and fault-tolerance of the data, as well as

good performance, must also be maintained. Sierra achieves all

these through a set of techniques including power-aware layout,

predictive gear scheduling, and a replicated short-term versioned

store. Replaying live server traces from a large e-mail service

(Hotmail) shows power savings of at least 23%, and analysis of

load from a small enterprise shows that power savings of up to 60%

are possible.

Bio: I've broad interests in systems. Currently I've been focusing

on file systems and storage technologies and high-performance data

centers. I also have great interest in applying machine learning

and queuing analysis to help simplify and automate system

management. Since September 2007 I have been a Researcher at

Microsoft Research in Cambridge, UK. I received my PhD/MS/BS from

Carnegie Mellon University.

16:15-17:00: Java performance analysis pitfalls

Speaker: Andy Georges (Ghent university)

Abstract: Performance measurements in a managed runtime system are

susceptible to a number of pitfalls. In this talk I will mainly

focus on one pitfall that has been prevalent the last decade: the

fact that managed runtime systems -- think JVMs -- can exhibit

non-deterministic behaviour. Contemporary JVMs use either a

mixed-mode approach where an interpreter is combined with a JIT

compiler (e.g., Hotspot) or they use a two-level compilation

scheme, combining a baseline compiler with an optimising

compiler. Most VMs use timer-based sampling at some point to allow

detection of hot methods that are subsequently

(re-)optimised. This results is a significant variability in

execution time for Java applications. The literature of the past

decade reveals a number of approaches people use to deal with

this, but sadly, almost non use a rigorous statistical approach. I

will outline the approach we propose to use for doing rigorous

performance analysis under two scenario's: (i) a non-controlled

execution, and (ii) an execution where the compilation is

controlled during measurements.

Bio: Andy Georges studied Computer Science at Ghent University in

Belgium. He obtained his masters degree in 1999 (back then called

'licentiaat'). After graduating, he worked for two years on

numerical solutions for Sturm-Liouville problems. Subsequently, he

worked on the development of a (synchronisation) record-replay

system for Java, targeting embedded systems. He started his

PhD. research in 2003, taking a long a hard look at the

methodologies researchers employed for analysing performance of

Java applications. Along the way he developed an interest in

statistics, machine learning, and performance analysis and

prediction in general. In 2008 he obtained his PhD -- titled Three

Pitfalls in Java Performance Evaluation -- in computer science

from Ghent University after a successful defense. Since then he

has been a post-doctoral researcher for the Fund for Scientific

Research (FWO-Flanders), broadening his research interests to

performance analysis in system virtualisation. When he is not

working, he is enjoying family life and in his rare spare time, he

enjoys reading SF and fantasy of the better kind and fiddling

around in Haskell.

Wednesday December 16th 10:30-12:00 (room C669) ===============================================

10:30-11:15: Scaling Online Social Networks without Pains

Speaker: Vijay Erramilli (Telefonica Research)

Paper: http://netdb09.cis.upenn.edu/netdb09papers/netdb09-final3.pdf

Online Social Networks (OSN) face serious scalability challenges

due to their rapid growth and popularity. To address this issue we

present a novel approach to scale up OSN called One Hop

Replication (OHR). Our system combines partitioning and

replication in a middleware to transparently scale up a

centralized OSN design, and therefore, avoid the OSN application

to undergo the costly transition to a fully distributed system to

meet its scalability needs. OHR exploits some of the structural

characteristics of Social Networks: 1) most of the information is

one-hop away, and 2) the topology of the network of connections

among people displays a strong community structure. We evaluate

our system and its potential benefits and overheads using data

from real OSNs: w that OHR has the potential to provide

out-of-the-box transparent scalability while maintaining the

replication overhead costs in check.

Bio: Vijay Erramilli is a junior researcher at Telefonica

Research, Barcelona where he works on problems related to online

social networks, network economics and privacy issues. He

graduated from Boston University with a PhD in Computer Science in

December 2008 and joined Telefonica in Jan 2009.

11:15-12:00: LiteGreen: Saving Desktop Energy using Virtualization

Speaker: Pradeep Padala (University of Michigan)

To reduce energy wastage by idle desktop computers in enterprise

environments, the typical approach is to put a computer to sleep

during long idle periods (e.g., overnight), with a proxy employed

to reduce user disruption by maintaining the computer's network

presence at some minimal level. However, the Achilles heel of the

proxy-based approach is the inherent trade-off between the

functionality of maintaining network presence and the complexity

of application-specific customization.

In this presentation, I will talk about LiteGreen, a system to

save desktop energy by virtualizing the user's desktop computing

environment as a virtual machine (VM) and then migrating it

between the user's physical desktop machine and a VM server,

depending on whether the desktop computing environment is being

actively used or is idle. Thus, the user's desktop environment is

"always on," maintaining its network presence fully even when the

user's physical desktop machine is switched off and thereby saving

energy. This seamless operation allows LiteGreen to save energy

during short idle periods as well (e.g., coffee breaks), which is

shown to be significant according to our analysis of over 65,000

hours of data gathered from 120 desktop machines. We have

prototyped LiteGreen on the Microsoft Hyper-V hypervisor. Our

findings from a small-scale deployment of the system as well as

from laboratory experiments and simulation analysis are very

promising, with energy savings of 72-86% with LiteGreen compared

to 35% with existing Windows and manual power management.