|
All talks are held in
606 Soda Hall from
1:00 to 2:00pm, preceded by lunch from 12:30pm to 1:00pm, unless otherwise specified. On this site, you will find upcoming
talk announcements and abstracts from previous talks in Berkeley's database seminar.
You can also receive talk announcements on our mailing list. If you are interested,
please see the instructions for adding yourself to the mailing list.
Important Note: Spring Semesters are usually reserved for students talks. As these often contain only preliminary results, and are not necessarily of immediate interest to a broad audience, announcements will not be send to the mailing list during that time. Announcements will still be send on the occasion of a visiting speaker, or when the speaker wishes for the talk to be publicized to a larger audience. The seminar webpage will contain information on the current schedule even when announcements are not sent, and visitors are still welcome to attend if they wish.
|
» Fall 2009 Calendar
| 03/05 |
Usher and Probabilistic Declarative Information Extraction |
More Info |
Kuang Chen & Daisy Zhe Wang |
|
|
» Abstract
ICDE practice talks
| 03/12 |
Cloud Programming: From Doom and Gloom to BOOM and Bloom |
More Info |
Neil Conway |
|
|
» Abstract
Datalog workshop practice talk
| 04/30 |
Why No and Why So? Causes and Responsibility for Query Answers and Non-Answers |
More Info |
» Abstract
We propose causality as a unified framework to explain query answers and non-answers, thus generalizing and extending several previously proposed approaches of provenance and missing query result explanations. Albeit related to provenance, causality is a more subtle notion, that consists of only a small fragment of the lineage. Our framework is based on the well studied definition of actual causes by Halpern and Pearl, which we transform to a database setting. This talk revolves around various issues and challenges that we encountered, ranging from effective model choices to complexity analysis. I will further talk about current ideas and possible directions that we are exploring.
| 05/07 |
Dissertation Talk: An Experiment in Declarativity |
More Info |
Tyson Condie |
UC Berkeley |
|
» Abstract
As the environment in which systems are deployed becomes increasingly complex, distributed and heterogeneous, traditional techniques for building system infrastructure have become increasingly expensive and error-prone. In this talk, I report on our experiences with trying to raise the level of abstraction at which systems software is implemented. In the future, we think that optimizing for programmer productivity and overall total cost of ownership will be more important than the traditional metrics of raw performance and resource consumption. We evaluated this hypothesis through an experiment in building "real system" software in a declarative language. The goal of this experiment was twofold. First, we wanted to determine whether the performance benefits offered by low-level languages really matters today. That is, we would like to determine if modern hardware enables "real systems" to be implemented in a very high-level language. Second, we wanted to identify the class of problems that declarative languages are good at solving. The declarative language that we used for this experiment is Datalog. I will describe how we used Datalog to implement three different types of system problems -- network protocols, query optimization techniques, and MapReduce scheduling. A section is devoted to each of these problems and I will conclude each with a discussion on what did and did not work well.
» Previous Semester Schedules:
» To subscribe to the mailing list:
To subscribe to a list, send e-mail to
majordomo@db.cs.berkeley.edu,
with a message (not a Subject line!) containing only the words:
subscribe <list name>
As an example, one database wannabe might send the message:
To: majordomo@db.cs.berkeley.edu
From: turing@acm.org (Alan Turing)
Subject: I wannabe!
subscribe dblunch
Unsubscribing is similar:
To: majordomo@db.cs.berkeley.edu
From: turing@acm.org (Alan Turing)
Subject: can't make it any more
unsubscribe dblunch
» To subscribe to the mailing list:
To subscribe to a list, send e-mail to
majordomo@db.cs.berkeley.edu,
with a message (not a Subject line!) containing only the words:
subscribe <list name>
As an example, one database wannabe might send the message:
To: majordomo@db.cs.berkeley.edu
From: turing@acm.org (Alan Turing)
Subject: I wannabe!
subscribe dblunch
Unsubscribing is similar:
To: majordomo@db.cs.berkeley.edu
From: turing@acm.org (Alan Turing)
Subject: can't make it any more
unsubscribe dblunch