Select a calendar:
Filter February Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for February 24, 2014
-
CS Colloquium: David Chu (Microsoft Research) - Surmounting two challenges of cloud gaming for mobile devices: network latency and server multi-tenancy
Mon, Feb 24, 2014 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: David Chu, Microsoft Research
Talk Title: Surmounting two challenges of cloud gaming for mobile devices: network latency and server multi-tenancy
Series: CS Colloquium
Abstract: Gaming on mobile devices is very popular. Cloud gaming such as Sony PlayStation's Now -- where remote servers perform game execution and rendering on behalf of thin clients that simply send input and display output frames -- appears to be well-suited for mobile devices, promising any device the ability to play any game any time. However, cloud gaming must confront network latency and server multi-tenancy. This talk introduces these two challenges, and our two respective solutions, DeLorean and DeeJay.
For latency, wireless network round trip times (RTTs) often exceed thresholds above which gamers find responsiveness acceptable. We present DeLorean, a speculative execution system for mobile cloud gaming that is able to mask latency. DeLorean produces speculative rendered frames of future possible outcomes, delivering them to the client one entire RTT ahead of time; clients perceive little latency. To achieve this, DeLorean combines: 1) future input prediction; 2) state space subsampling and time shifting; 3) misprediction compensation; and 4) bandwidth compression. This work is a collaboration with the University of Michigan.
For multi-tenancy, a single server must carefully schedule the GPU across multiple game instances that each have their own real-time latency and throughput requirements. Moreover, it must gracefully handle overload when more clients join than anticipated. We are in the process of building DeeJay, a system that 1) schedules GPU-bound jobs with latency and throughput constraints, and that 2) minimally degrades visual game quality upon system overload.
To evaluate both DeLorean and DeeJay, we use two high quality, commercially-released games: a twitch-based first person shooter, Doom3, and a role playing game, Fable3.
Biography: David Chu is a researcher at Microsoft Research in Redmond where he works on mobile systems with an emphasis on mobile gaming. He is also interested in sensing and context for the mobile OS. His work has appeared in The Verge, Engadget and Wired. David received his Ph.D. and M.S. from the University of California, Berkeley, and his B.S. from the University of Virginia.
Host: Ramesh Govindan
Location: SAL 222
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
PhD Defense - Sumita Barahmand
Mon, Feb 24, 2014 @ 01:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Benchmarking Interactive Social Networking Actions
PhD Candidate: Sumita Barahmand
Defense Committee: Shahram Ghandeharizadeh (Chair), Ramesh Govindan, Nenad Medvidović and Bhaskar Krishnamachari (Outside Member)
Date: Monday, February 24, 2014
Time: 1:00 PM
Location: EEB 248
Abstract:
Social networking sites such as Google+, Facebook, Twitter and LinkedIn, are cloud service providers for person to person communications. There are different approaches to building these sites ranging from SQL to NoSQL and NewSQL, Cache Augmented SQL, graph databases and others. Some provide a tabular representation of data while others offer alternative models that scale out. Some may sacrifice strict ACID (Atomicity, Consistency, Isolation, Durability) properties and opt for BASE (Basically Available, Soft-state, Eventual consistency) to enhance performance. Independent of a qualitative discussion of these approaches and their merits, a key question is how do these systems compare with one another quantitatively? This dissertation investigates the viability of a benchmark to address this question.
Our primary contribution is the design and implementation of a novel benchmark for interactive social networking actions named BG. BG's design decisions are as follows: First, it rates the performance of a system for processing interactive social networking actions by computing two values: Socialites and Social Action Rating (SoAR) using a pre-specified SLA. An example SLA may require 95\% of issued requests to observe a response time faster than 100 milliseconds. Second, BG elevates the amount of unpredictable data produced by a solution to a first class metric, including it as a key component of the SLA (similar to the average response time) and quantifying it as a part of the benchmarking process. It also computes the freshness confidence to characterize the behavior of a weak consistency technique. Third, BG's generated workload is characterized by reads and writes of a very small amount of data from big data. Fourth, BG is a modular, extensible framework that is agnostic to its underlying data store. Fifth, BG employs a logical partitioning of data to scale both vertically and horizontally to thousands of nodes. This is essential for evaluating scalable installations consisting of thousands of nodes. Finally, BG includes a visualization tool to empower an evaluator to monitor an in-progress benchmark and identify bottlenecks.
BG's possible use cases are diverse. One may use BG to compare and contrast various data stores with one another, characterize tradeoffs associated with alternative physical representations of data, or quantify the behavior of a data store in the presence of various failures (either CP or AP of the CAP theorem) among the others. This dissertation demonstrates use of BG in two contexts. First, to rate an industrial strength relational database management system and a document store, quantifying their performance tradeoffs. This analysis includes the use of a middle tier cache (memcached) and its impact on the performance of each system. Second, to gain insight into alternative design decisions for implementing a social action by characterizing their behavior with different social graphs and system loads. BG's proposed framework is quite novel and opens several new research directions that benefit the systems research community.
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
Alexander V. Terekhov: Constructing space: how a naive agent can learn spatial relationships by observing sensorimotor contingencies
Mon, Feb 24, 2014 @ 01:00 PM - 02:30 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Alexander V. Terekhov, Laboratory of Psychology of Perception, Paris Descartes University (Paris 5).
Talk Title: Constructing space: how a naive agent can learn spatial relationships by observing sensorimotor contingencies
Series: CS Colloquium
Abstract: The brain sitting inside its bony cavity sends and receives myriads of sensory inputs and outputs. A problem that must be solved either in ontogeny or phylogeny is how to extract the particular characteristics within this "blooming buzzing confusion" that signal the existence and nature of physical space, with structured objects immersed in it, among them the agent's body. The idea that spatial knowledge must be extracted from the sensorimotor flow in order to underlie perception has been considered by a number of thinkers, including Helmholtz, Poincare, Nicod, Gibson, etc. However, little work has considered how this could actually be done by organisms without a priori knowledge of the nature of their sensors and effectors. Here we show how an agent with arbitrary sensors will naturally discover spatial knowledge from the undifferentiated sensorimotor flow. The method first involves tabulating sensorimotor contingencies, that is, the laws linking sensory and motor variables. Second, further laws are created linking these sensorimotor contingencies together. The method works without any prior knowledge about the structure of the agent's sensors, body, or of the world. We show that the extracted laws endow the agent with basic spatial knowledge, manifesting itself through perceptual shape constancy and the ability to do path integration. We further show that the ability of the agent to learn all spatial dimensions depends on the ability to move in all these dimensions, rather than on possessing a sensor that has that dimensionality. This latter result suggests, for example, that three dimensional space can be learned in spite of the fact that the retinas are two-dimensional. We conclude by showing how the acquired spatial knowledge paves the way to building the notion of object.
Host: Michael Arbib
Location: HNB 100
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
