2012 Meetings

Date: Wednesday, November 7th, 2012

Opportunities and Challenges in Performance and Resource Management in the Cloud

Dr. Xiaoyun Zhu, VMware Inc.

Date: Wednesday, November 7th, 2012, 6:45-8:00pm
Location: Carnegie Mellon Silicon Valley, Building 23,Moffett Field, CA, Room 118
Registration (for headcount): http://ieeeciscloud.eventbrite.com

Abstract: In the past decade, the information technology industry has experienced a paradigm shift as computing resources became available to businesses and individuals as a utility through cloud based services. The wide adoption of server virtualization technologies has laid a foundation for both private and public cloud environments. The benefits of virtualized infrastructures are multifaceted, including easier deployment, elastic capacity, higher availability, higher resource utilization, and lower energy cost. At the same time, it brings many new challenges to the areas of application performance management and service level assurance. These challenges make it impossible for human administrators to carry out monitoring, detection, analysis, and remediation of performance problems on a 24x7 basis. On the other hand, they present unique opportunities in applying control theory, optimization, and statistical learning based techniques to developing model-based, automated performance diagnosis and control frameworks. There has been some amount of success in this area in the last several years, but many problems remain. In addition, much of the research work published was evaluated on lab test beds of limited scale, with a small number of applications and virtual machines. On the other hand, real cloud environments today can easily have thousands or more physical hosts and hundreds of thousands of VMs. So any performance management automation solution developed will have limited practicality unless it passes the scalability test. In this talk, I’ll highlight some of the open research problems and related technical challenges, in hope to attract more invocative ideas and solutions from a larger community of researchers and practitioners.

Biography: Xiaoyun Zhu is a Staff Engineer in the VMware Cloud Resource Management Group, focusing on developing automated resource and performance management solutions for virtualized datacenters and applications. Her general interests are in applying control theory, optimization, algorithms, and statistical learning to IT systems management and automation. Prior to VMware, she was a Senior Research Scientist at Hewlett Packard Labs for 8 years. She has co-authored over 50 refereed papers in journals and conference proceedings, and holds 17 patents. She has been a program committee member for ICAC, IM, NOMS, CNSM, MASCOTS, CCGrid, and SIGMETRICS, and an associated editor for the Journal of Network & Systems Management. Xiaoyun co-founded the first International FeBID Workshop (now known as Feedback Computing Workshop) in 2006, along with Joe Hellerstein and Tarek Abdelzaher, and has been on the steering committee of FeBID since then. She will be the Program Chair for ICAC in 2013. Xiaoyun received her dual B.S. in Automation and Applied Mathematics from Tsinghua University in 1994, and her Ph.D. in Electrical Engineering from California Institute of Technology in 2000.

Directions: Exit 101 on Moffett Blvd Exit and proceed to the Security Gate. You will pass through the security gate to reach the campus. A valid driver’s license, state ID, or federal ID is required. When presenting your ID, let them know you are going to Carnegie Mellon.

After Passing Through the Security Gate: Once you proceed through security gate, move into the right lane before you approach the “Y” in the road (about 200 yards from the security gate). Continue in the right lane as the road bends right, a chapel will be on your right. Carnegie Mellon’s Building 23 is directly to your left. Make an immediate left into the Building 23 parking lot. You are now in the back of Building 23. Follow the signs/lights to the main entrance in the front of the building.

The meeting will start with food and beverages at 6:45 PM with the talk starting at 7:00PM.


Date: Wednesday, November 28th, 2012

Co-sponsored event with SCV IEEE Control Systems Society:

The Passage through Resonance of a Coupled Mechanical Oscillator - Part 1: The Experiment

Dr. Ghulam Mustafa

Date: Wednesday, November 28th, 2012, 6:30-8:00pm

Location: Intersil Corp., 1001 Murphy Ranch Road, Milpitas, CA 95035 (Building 3, Stanford Room)

Registration (for headcount): http://css-ghulam2012-part1.eventbrite.com/


Date: Wednesday, September 26, 2012

Applications of Extreme Value Theory to Signal Processing

Dr. Adam Rowell, Stanford University

Date: Wednesday, September 26, 2012
7:00pm: Welcome, Introductions
7:10pm: Presentation
8:00pm: Q&A, Networking

Location: Packard 101 at Stanford University from 7:00 to 8:30PM.
Directions here. Parking is free in the evening in the large parking structure at the corner of Via Ortega and Panama St.

Registration (for headcount): http://cis-rowell-evt.eventbrite.com/

Co-Sponsors: SCV IEEE Signal Processing Society, SCV IEEE Circuits & Systems Society, SCV IEEE Control Systems Society

Abstract: Extreme value theory (EVT) is the study of the statistics of the extreme outliers in a random process. It is especially useful for estimating probabilities of an extreme event when little or no past data has been recorded at a similar level. Canonical examples of its application include predicting annual maximum river or wave heights and estimating worst-case insurance or stock market losses. Many problems in electrical engineering can benefit from the application of EVT, though such research is just getting started. In this talk, we will go over the basics of extreme value theory and show how its principles are useful to a few signal processing applications.

Both examples we investigate will illustrate how existing problems in electrical engineering can be easily tackled using EVT, yielding useful models. For the first application, we investigate the effects of quantization on digital filter performance. When digital filter coefficients are quantized, as is common in high-speed or low-power hardware, the performance can be significantly degraded. We will start with a quick overview of simple digital filters and their frequency response, and will see the effects of quantization on performance. Extreme value theory will then be applied to the frequency response of the quantized filter to model how the quantization affects the maximum response error. In our second example, we will use extreme value theory to model overflow rates in digital systems. Even without knowing the underlying distributions of the data being analyzed, EVT can accurately estimate the rate at which a value will exceed a high threshold.


Dr. Adam Rowell
Adam Rowell is currently finishing his PhD in Electrical Engineering at Stanford University under Dr. Bernard Widrow. His research focus is using Extreme Value Theory to solve signal processing problems, including optimizing the performance of quantized digital filters, and estimating overflow rates of fixed-point digital systems. Past research has also involved studying adaptive signal processing algorithms, and working on classifying different types of epilepsy in children using EEG signals. He has previously worked for the Signal Processing Toolbox and Fixed-Point Toolbox teams at The MathWorks Corporation, where he worked on tools for converting digital systems from a floating-point design to a fixed-point implementation. After completing his PhD in the fall of 2012, he will begin working at Exponent Consulting, doing failure analysis and rapid prototyping in the Electrical Engineering and Computer Science division, with emphasis on signal processing applications. Adam can be contacted via email at 'rowell [at] stanford [dot] edu'.


Date: Wednesday, September 19, 2012
Main Sponsor: SCV IEEE Control Systems Society

Control of the Disk Drive Actuator – 1980 and Today

Art Wagner and Dick Oswald

Date: Wednesday, September 19, 2012, 6:30-8:00pm
Location: Intersil Corp., 1001 Murphy Ranch Road, Building 3/Stanford Room, Milpitas, CA 95035.
Registration (for headcount): http://ieeecss-wagner.eventbrite.com/

Abstract: In computer systems, an important data storage medium, beginning in the 1950s until today 2012, is the Hard Disk Drive (HDD). Using magnetic heads, the HDD reads and writes data on magnetic surfaces of spinning disks. A control system positions the magnetic heads using a disk drive actuator that is composed of magnets, steel, a coil, bearings, and head arms. The two main control modes are “seek” and “track follow”. In this presentation we use the disk drive actuator and control modes prevalent in c. 1980 as a reference design for comparison with today’s HDD.

Biography: Art taught full time at SJSU for 13 years attaining full professorship with tenure, then he went into the disk drive industry. He designed in the areas of the magnetics and control of the actuator and the spindle motor for a multitude of disk drive companies, including Seagate, Maxtor, Maxoptics, Quantum, Conner Peripherals, IBM, ISS, Priam, Iota, StorCard, and Swan. Along the way, Art taught a series of short courses on the moving coil actuator, the disk drive spindle motor, and perpendicular magnetic recording. He also taught classes part-time at Santa Clara University on mechatronics. Presently, he is teaching a class at SJSU. Art received a bachelor’s degree from Santa Clara University, master’s degree from the University of Arizona, and a Ph.D. from Oregon State University.


Date: Thursday, January 26, 2012

Quantization Noise

Prof. Bernard Widrow, Stanford University

Date: Thursday, January 26, 2012, 7:00-9:00pm
Location: Packard Electrical Engineering Building, Room 101, 350 Serra Mall, Stanford, CA 94305

Abstract: For years, rumors have been circulating about quantization noise:
(a) The noise is additive and white and uncorrelated with the signal being quantized, and
(b) The noise is uniformly distributed between plus and minus half a quanta, having zero mean and a mean square of one twelveth the square of a quanta.

Yet, simple reasoning tells another story:
(a) The noise is related to the signal being quantized,
(b) The probability distribution of the noise depends on the probability distribution of the signal being quantized, and
(c) The noise will be correlated over time if the signal being quantized is correlated over time.

In spite of the simple reasoning, the rumors turn out to be true when quantizing theorems, analogous to the well known sampling theorem, are satisfied. Quantization, a nonlinear process, can be analyzed by linear sampling theory applied to the probability density distribution of the signal being quantized. In practice, the quantizing theorems are almost never perfectly satisfied (by analogy, signals are almost never perfectly band-limited so the sampling theorem is almost never perfectly satisfied). However, the rumors about quantization noise turn out to be extremely close to being true for a wide practical range of signal characteristics.

When the rumors are true, signal processing, communication, and control systems containing nonlinear quantization behave like noisy linear systems and are easy to analyze. The original theory was developed by B. Widrow in his 1956 MIT doctoral thesis, applied to fixed-point (uniform) quantization. The theory was extended by Widrow and I. Kollar in the 1990's to apply to floating-point quantization. Their work on these subjects was published in a Cambridge University Press book in 2008 entitled "Quantization Noise".

Biography: Dr. Bernard Widrow (M'56, SM'70, F'76, LF'06) received from MIT the S.B. degree in 1951, the S.M. degree in 1953, and the Sc.D. degree in 1956, all in electrical engineering. At MIT from 1956 -1959, he taught classes in digital signal processing and started research in what was to become the field of adaptive signal processing. In 1959, he joined the faculty at Stanford University.

At Stanford, he introduced graduate courses in digital signal processing, adaptive signal processing, and artificial neural networks. During his 50 years at Stanford, he was the principal advisor to 85 Ph.D. students. His students have become workers in industry, founders of companies, professors, deans, provosts, medical doctors, and admirals in the U.S. Navy.

In 1959, together with his first Ph.D. student, M.E. Hoff, Jr., he invented the LMS algorithm (least mean square), still considered to be the world's most widely used learning algorithm. It is used in all modems for channel equalization and echo cancelling. It is one of the key technologies that make the internet possible. LMS is used in adaptive neural networks, controls, noise cancelling, antenna arrays, speech recognition, face recognition, devices for the hearing impaired, and in other medical applications.

This research has led to several hundred journal publications, twenty patents, and four books. One of his papers, "Adaptive Antenna Systems", published in December 1967 in the Proceedings of the IEEE, became a citation classic. His four books are B.Widrow and S.D.Stearns, "Adaptive Signal Processing", Prentice-Hall, 1985; B.Widrow and E.Walach, "Adaptive Inverse Control", Prentice-Hall, 1996; S. Haykin and B.Widrow, eds.,"Least-Mean-Square Adaptive Filters", Wiley-Interscience, 2003; B.Widrow and I.Kollar,"Quantization Noise", Cambridge University Press, 2008.

Dr. Widrow is a Life Fellow of the IEEE and a Fellow of AAAS. He received the IEEE Centennial Medal in 1984, the IEEE Alexander Graham Bell Medal in 1986, the IEEE Neural Networks Pioneer Medal in 1991, the IEEE Signal Processing Society Medal in 1998, the IEEE Millennium Medal in 2000, and the Benjamin Franklin Medal from the Franklin Institute in 2001. He was inducted into the National Academy of Engineering in 1995, and into the Silicon Valley Engineering Hall of Fame in 1999. He is currently Chairman of the Santa Clara Valley Chapter of the IEEE Computational Intelligence Society.