Past Events
9 February
K2I Distinguished Lecture
Speaker: Eric Horvitz
Thursday, February 9, 2012
4:00 PM to 5:00 PM
Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Thursday, February 9, 2012
4:00 PM to 5:00 PM
Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
8 February
TBA
Speaker: Dora Angelaki
Wednesday, February 8, 2012
4:00 PM to 5:00 PM
1049 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Host: Tony Elam
Wednesday, February 8, 2012
4:00 PM to 5:00 PM
1049 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Host: Tony Elam
3 February
Nested QoS: Providing Flexible SLAs in Shared Storage Systems
Speaker: Hui Wang
Friday, February 3, 2012
10:00 AM to 11:00 AM
3110 Duncan Hall
The increasing popularity of storage and server consolidation introduces new challenges for resource management, capacity provisioning, and guaranteeing application performance. Typical Service Level Agreements (SLAs) provide performance guarantees in terms of throughput (IOPS) or response time limits (ms). The bursty nature of storage workloads [1] (where instantaneous arrival rates significantly exceed the average) implies a large gap between peak and average resource requirements in meeting response time bounds, leading to low overall server utilization and high cost. This situation is driving the development of elastic QoS models that allow clients greater flexibility in adopting SLAs tailored to their workload characteristics and performance requirements, while allowing the service provider opportunities to optimize provisioning and scheduling decisions. In this thesis we present a novel Nested QoS service model for multiplexing multiple concurrent bursty workloads on a shared storage server. Our solution employs two strategies together: systematically classifying requests to provide each workload with a graduated Quality of Service (QoS) and flexibly scheduling the classified portions of all the workloads. The results show that the Nested QoS service model provides (i) performance isolation and strong performance guarantees in spite of the badly behaved workload (2) a flexible and auditable elastic SLA definition (3) improved server utilization.
Friday, February 3, 2012
10:00 AM to 11:00 AM
3110 Duncan Hall
The increasing popularity of storage and server consolidation introduces new challenges for resource management, capacity provisioning, and guaranteeing application performance. Typical Service Level Agreements (SLAs) provide performance guarantees in terms of throughput (IOPS) or response time limits (ms). The bursty nature of storage workloads [1] (where instantaneous arrival rates significantly exceed the average) implies a large gap between peak and average resource requirements in meeting response time bounds, leading to low overall server utilization and high cost. This situation is driving the development of elastic QoS models that allow clients greater flexibility in adopting SLAs tailored to their workload characteristics and performance requirements, while allowing the service provider opportunities to optimize provisioning and scheduling decisions. In this thesis we present a novel Nested QoS service model for multiplexing multiple concurrent bursty workloads on a shared storage server. Our solution employs two strategies together: systematically classifying requests to provide each workload with a graduated Quality of Service (QoS) and flexibly scheduling the classified portions of all the workloads. The results show that the Nested QoS service model provides (i) performance isolation and strong performance guarantees in spite of the badly behaved workload (2) a flexible and auditable elastic SLA definition (3) improved server utilization.
31 January
ECE Distinguished Lecture
EM2: Execution Migration Machine
Speaker: Srini Devadas
Tuesday, January 31, 2012
4:00 PM to 5:00 PM
TBA Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
We introduce the Execution Migration Machine (EM2) a computation-migration-based multicore architecture that provides speedy access to on-chip distributed cache data by either migrating execution or via remote memory operations. Since only one copy of data is stored on-chip in a Non Uniform Cache Access (NUCA)-style organization, cache coherence and sequential consistency are trivially ensured without the need for coherence logic or large directories. We develop a one-step migration protocol that is deadlock-free based on the concept of cores native to a thread, and present migration prediction algorithms under this protocol that decide when to migrate or otherwise perform a remote access, and decide what part of the context to carry during a migration. EM2 performs better than conventional remote access in a NUCA organization, because it better exploits locality.
Tuesday, January 31, 2012
4:00 PM to 5:00 PM
TBA Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
We introduce the Execution Migration Machine (EM2) a computation-migration-based multicore architecture that provides speedy access to on-chip distributed cache data by either migrating execution or via remote memory operations. Since only one copy of data is stored on-chip in a Non Uniform Cache Access (NUCA)-style organization, cache coherence and sequential consistency are trivially ensured without the need for coherence logic or large directories. We develop a one-step migration protocol that is deadlock-free based on the concept of cores native to a thread, and present migration prediction algorithms under this protocol that decide when to migrate or otherwise perform a remote access, and decide what part of the context to carry during a migration. EM2 performs better than conventional remote access in a NUCA organization, because it better exploits locality.
Host: Farinaz Koushanfar
30 January
The Lasso: some novel algorithms and applications
Speaker: Rob Tibshirani
Monday, January 30, 2012
4:00 PM to 5:00 PM
1070 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Dr. Tibshirani will discuss some procedures for modelling high-dimensional data, based on L1 (lasso) style penalties. He will describe pathwise coordinate descent algorithms for the lasso, which are remarkably fast and facilitate application of the methods to very large datasets for the first time. He will then give examples of new applications of L1 penalties to microarray classification, the fused lasso for signal detection and the matrix completion problem.
Monday, January 30, 2012
4:00 PM to 5:00 PM
1070 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Dr. Tibshirani will discuss some procedures for modelling high-dimensional data, based on L1 (lasso) style penalties. He will describe pathwise coordinate descent algorithms for the lasso, which are remarkably fast and facilitate application of the methods to very large datasets for the first time. He will then give examples of new applications of L1 penalties to microarray classification, the fused lasso for signal detection and the matrix completion problem.
Bio: Robert Tibshirani is a Professor in the Departments of Statistics and Health Research and Policy at Stanford University. He was a Professor at the University of Toronto from 1985 to 1998. In his work, he develops statistical tools for the analysis of complex datasets, most recently in genomics and proteomics. His most well-known contributions are the lasso , which uses L1 penalization in regression and related problems, and Significance Analysis of Microarrays (SAM) . He also co-authored three widely used books "Generalized Additive Models", "An Introduction to the Bootstrap", and "The Elements of Statistical Learning".
*In recognition of The Department of Statistics 25th year celebration, there will be a reception following this talk in Martel Hall.
26 January
Time and the Brain
Speaker: David Eagleman
Thursday, January 26, 2012
4:00 PM to 5:00 PM
1070 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Most of the actions our brains perform on a daily basis -- such as perceiving, speaking, and driving a car -- require timing on the scale of tens to hundreds of milliseconds. New discoveries in our laboratory are contributing to an emerging picture of how the brain processes, learns, and perceives time. We will demonstrate new temporal illusions in which durations dilate, perceived order of actions and events are reversed, and time is experienced in slow motion. Questions addressed include: Does your brain work in real time, or do you experience a delayed version of the world? How and why does the brain dynamically recalibrate its timing judgments? Does subjective time really slow down during a car accident?
Thursday, January 26, 2012
4:00 PM to 5:00 PM
1070 Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Most of the actions our brains perform on a daily basis -- such as perceiving, speaking, and driving a car -- require timing on the scale of tens to hundreds of milliseconds. New discoveries in our laboratory are contributing to an emerging picture of how the brain processes, learns, and perceives time. We will demonstrate new temporal illusions in which durations dilate, perceived order of actions and events are reversed, and time is experienced in slow motion. Questions addressed include: Does your brain work in real time, or do you experience a delayed version of the world? How and why does the brain dynamically recalibrate its timing judgments? Does subjective time really slow down during a car accident?
Host: Tony Elam
24 January
K2I Distinguished Lecture: Limor Fix
Speaker: Limor Fix
Tuesday, January 24, 2012
4:00 PM to 5:00 PM
Rice University
6100 Main St
Houston,Texas,USA
Tuesday, January 24, 2012
4:00 PM to 5:00 PM
Rice University
6100 Main St
Houston,Texas,USA
20 January
High performance quantum cascade lasers in the 3-5 um spectral range
Electrical and Computer Engineering
Digital Signal Processing
Friday, January 20, 2012
11:00 AM to 12:00 PM
202 Space Science Building
Rice University
6100 Main St
Houston,Texas,USA
Recently, the need for compact sensors, capable of real-time concentration monitoring, has become crucial. For this reason nowadays a strong interest is shown in the optimization of low-consumption quantum cascade lasers (QCLs) and in particular on the improvement of their power efficiency. QCL performance is strongly dependent on the careful optimization of the active region design. Therefore a reliable simulation model is important to predict laser behaviour. In our group a new transport model has been developed[1] that can not only model intraperiod electron dynamics, e.g. scattering times, but it is also able to estimate actual laser performance, i.e. the output power and the voltage-current characteristic. This method enormously eases the route to an efficient and fast optimization of the laser design. One of the key elements to be addressed in designing a QCL structure is the optimization of the injector efficiency in to the upper lasing state. This is particularly crucial for lasers in the short wavelength range (3-5 µm). In this case the high electric field needed to align the injector ground state with the upper lasing transition, leads to relevant parasitic currents flowing in the structure already before the levels align. To reduce these leakage currents, while maintaining an efficient coupling of the injector state with the upper lasing state, new designs based on composite injector barriers (Step Well designs) will be introduced. This design show a relevant improvement of the laser efficiency.
Among the first atmospheric window, the 3-4 µm spectral region is particularly interesting for industrial applications since the fundamental C-H, N-H and O-H stretching modes have strong resonances in this region(e.g. Methane, Formaldehyde). Unfortunately for wavelengths smaller than 4 µm, the realization of QCLs is challenging using the standard material systems due to the large conduction band discontinuity needed.
The development of Sb-free QCL designs based on the InGaAs/AlInAs-AlAs on InP material system emitting in the 3-4 µm range will be presented. Maximum operation temperature above 350K will be presented at 3.3 µm with watt-level performance[2]. Previously, no QCL with emission wavelength smaller than 3.6 µm was demonstrated in this material system in a temperature accessible by thermoelectric cooling[3]. Moreover, using an external cavity configuration, spectral tuning over more than 275 cm-1 has been observed in pulsed operation.
The use of this material system(InGaAs/AlInAs-AlAs) is particularly appealing since it is the only one Sb-free in this spectral range and it is fully compatible with the high-performance QCL fabrication processes for continuous wave emission, i.e. buried heterostructure process(BH). Preliminary results on the distributed feedback BH lasers will also be presented. Single mode emission at room temperature will be presented.
[1] R. Terazzi, and J. Faist, New Journal of Physics 12, 033045 (2010).
[2] A.Bismuto,M.Beck,J.Faist Appl.Phys.Lett.98,191104(2011).
[3] J.Faist,F.Capasso,D.Sivco,A.Hutchinson,S.Chu, A.Cho Appl.Phys.Lett.72,680(1998).
Host: Frank K. Tittel (AL A237 x4833)
19 January
TBA
Speaker: Parmesh Ramanathan
Thursday, January 19, 2012
4:00 PM to 5:00 PM
TBA Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Host: Lin Zhong
Thursday, January 19, 2012
4:00 PM to 5:00 PM
TBA Duncan Hall
Rice University
6100 Main St
Houston,Texas,USA
Host: Lin Zhong
9 January
Detection schemes, algorithms and device modeling for TDLS
Speaker: Andreas Hangauer
Monday, January 9, 2012
11:00 AM to 12:00 PM
202 Space Science Building
Rice University
6100 Main St
Houston,Texas,USA
The basic question which provides the frame of the work is what TDLS sensor
properties can be improved when a digital sensor hardware is utilized.
One topic I will present is the analysis of multi-harmonic detection schemes
for wavelength modulation spectroscopy. I will show theroretically and
experimentally that multi-harmonic detection has advantages in terms of
noise over standard second harmonic detection and will outline the method
how to implement the proper data processing and anaysis. In course of this
I will also show general properties of harmonic spectra including an
improved computation method for these including their derivatives. This is
important for spectral least squares curve-fitting using the Voigt or Gauss
absorption line profile because no numerical integration is required. It is
based on spline approximation of the transmission spectrum and an exact
closed form expression for the harmonic spectrum of a spline function.
The second major topic is device modeling with aim of creating computer
simulations of TDLS sensors to improve algorithms and detection schemes in
case this is not possible analytically. The challenge for this is proper
modeling of the laser diode in the lower frequency range (< 100 MHz)
including thermal tuning and nonlinear intensity modulation behavior. This
is not done in literature so far since these effects are negligible for
telecomunication applications. I will present a general theory and
comparison with experimental results for VCSELs of different types and
operating wavelengths. This significantly improves the understanding and
modeling of the tuning behavior of these lasers.
Monday, January 9, 2012
11:00 AM to 12:00 PM
202 Space Science Building
Rice University
6100 Main St
Houston,Texas,USA
The basic question which provides the frame of the work is what TDLS sensor
properties can be improved when a digital sensor hardware is utilized.
One topic I will present is the analysis of multi-harmonic detection schemes
for wavelength modulation spectroscopy. I will show theroretically and
experimentally that multi-harmonic detection has advantages in terms of
noise over standard second harmonic detection and will outline the method
how to implement the proper data processing and anaysis. In course of this
I will also show general properties of harmonic spectra including an
improved computation method for these including their derivatives. This is
important for spectral least squares curve-fitting using the Voigt or Gauss
absorption line profile because no numerical integration is required. It is
based on spline approximation of the transmission spectrum and an exact
closed form expression for the harmonic spectrum of a spline function.
The second major topic is device modeling with aim of creating computer
simulations of TDLS sensors to improve algorithms and detection schemes in
case this is not possible analytically. The challenge for this is proper
modeling of the laser diode in the lower frequency range (< 100 MHz)
including thermal tuning and nonlinear intensity modulation behavior. This
is not done in literature so far since these effects are negligible for
telecomunication applications. I will present a general theory and
comparison with experimental results for VCSELs of different types and
operating wavelengths. This significantly improves the understanding and
modeling of the tuning behavior of these lasers.
Host: Frank K. Tittel (AL A237, x4833)