CSE 707 Seminar: Select Topics on Modern Database Systems (Fall 2022)

Modern database systems have different designs compared to traditional RDBMS in many aspects. In this seminar, we will review and discuss a range of papers of modern database system designs. The topics will include query processing, transaction processing, indexing and storage, and etc.
Prerequisites: CSE 462/562 Database Systems or equivalents.

Logistics

Location and Time: Davis Hall 113A, Wednesday 10:00 am to 12:50 pm.
Instructor: Zhuoyue Zhao, zzhao35 [at] buffalo [dot] edu, Davis Hall 338I.
Office hours are on demand. Email me for an appointment.
No required textbook.
Optional readings: Readings in Database Systems, 5th Edition, by Peter Bailis, Joseph M. Heller- stein, and Michael Stonebraker. Available online.
Attendance is required (see grading policy).
We'll be using Piazza for discussion and Q&A. Find our class page here.
Please make all required assignment submissions to UBLearns.

Course Requirements (Please Read)

• Read the required paper before coming to the lectures.
• List three questions before each lecture and submit them to UBLearns one day in advance (due on every Tuesday, 10:00 AM).
• Write a short paper summary after each lecture and submit it to UBLearns within two days (due on every Friday, 10:00 AM). The only exception is the weeks with two presentations, where the deadline will be extended to (Saturday, 11:00 PM).
• Present one of the selected papers. Your presentation will be graded on its soundness and completeness. Please submit your presentation slides before the lecture and no later than one day prior to your presentation (due Tuesday 10 AM). You may also be required to make changes to your presentation slides and resubmit it prior to the presentation Therefore, make sure you are reachable through Piazza before your presentation to avoid losing points. The presentation should be longer than a usual conference talk (plan for 1 hour presentation + 30 min discussion), and discuss the topics in depth, including a brief introduction to the background, the problem solved, the solution, as well as experimental evaluation if available in the paper. Note that the presentation slides may be based on existing works of others (e.g., the conference talk slides from the paper authors), as long as there are proper citations and acknowledgments. However, you should add your own contents in your presentation slides in most cases because 1) the slides need to reflect your own opinions; 2) conference talk slides usually do not cover enough technical details due to the time limit of conference talks.
  Some useful tips for presentation preparation: you are encouraged to start early on your presentation (e.g., one or two weeks in advance) if not familiar with the topics and you may have to spend a few days to go through the background and the technical details of the paper. You are also encouraged to discuss the paper with me and Feel free to post a private message on Piazza for any questions or for scheduling office hours.

Grading

• 8% for the pre-lecture questions (0.5% each).
• 32% for the paper summaries (2% each).
• 30% for the presentation.
• 30% for participation (evaluated based on your attendance, in-class discussion, and random quizzes).

The seminar is graded on S/U basis. Score >= 75%: satisfactory and score < 75: unsatisfactory.

Course Schedule

For UB students: Some of the following links may require an ACM Digital Library subscription. UB library has a paid subscription available for all students so you do not need to pay. You may either connect to eduroam when you're on campus, or replace dl.acm.org with dl-acm-org.gate.lib.buffalo.edu and enter your UBIT login credentials when you're off campus.

* Paper summary deadline is extended to Saturday at 11:00 PM for the weeks with two presentations.

The following schedule is subject to change due to student add/drop. Please double check the latest schedules when completing your assignments.

Date	Topic	#	Required Readings	Presenter
8/31/2022	Logistics and Introduction		Lecture Slides	N/A
Online Analytical Query Processing
9/7/2022	Columnar store	10	Andrew Lamb, et al. The Vertica Analytic Database: CStore 7 Years Later. In VLDB '12	Anirudh
9/14/2022	Vectorized query execution	20	Peter Boncz, et al. MonetDB/X100: Hyper-Pipelining Query Execution. In CIDR '05.	Rohit
9/21/2022	Query compilation	30	Thomas Neumann. Efficiently Compiling Efficient Query Plans for Modern Hardware. In VLDB '11.	Rahul
9/28/2022	SIMD	40	Orestis Polychroniou, et al. Rethinking SIMD Vectorization for In-Memory Databases. In SIGMOD '15.	Haohua
Approximate Query Processing
*10/5/2022 (Davis 338A)	Online Aggregation	50	Joseph M. Hellerstein, et al. Online aggregation. In SIGMOD '97.	Chaoping
	Overview of Approximate Query Processing	60	Chaudhuri et al. Approximate Query Processing: No Silver Bullet	Hon Ching
*10/12/2022	Random Sample Generation	70	Frank Olken. Random Sampling From Databases, Section 2.4 - 2.6; and Jeffery S. Vitter. Random Sampling with a Reservoir.	Roshan
		80	Frank Olken. Random Sampling From Databases, Section 2.7 and Section 3; and Alastair J. Walker. An Efficient Method for Generating Discrete Random Variables with General Distributions. In ACM TOMS'77.	Pratik
10/19/2022	Independent Range Sampling	90	Xiaocheng Hu, et al. Independent Range Sampling. In PODS '14.	Harish
10/26/2022	Join Sampling	100	Zhuoyue Zhao, et al. Random Sampling over Joins Revisited. In SIGMOD '18.	Karan
11/2/2022	Sample Set Selection	110	Sameer Agarwal, et al. BlinkDB: Queries with Bounded Errors and Bounded Response Times on Very Large Data. In Eurosys '13.	Gunjan
11/9/2022	Joins in AQP	120	Peter J. Haas, et al. Ripple Joins for Online Aggregation. In SIGMOD '99; and Chris Jermaine, et al. Scalable approximate query processing with the DBO engine. In SIGMOD '07;	Raman
11/16/2022	Joins in AQP (cont'd)	130	Feifei Li, et al. Wander Join: Online Aggregation via Random Walks. In SIGMOD '16.	Gang
11/23/2022	Fall recess, no lecture today.
11/30/2022	Joins in AQP (cont'd)	140	Yu Chen, et al. Two-Level Sampling for Join Size Estimation. In SIGMOD '17.	Sphoorthi
*12/7/2022	ML-based AQP (cont'd)	150	Qingzhi Ma, et al. DBEst: Revisiting Approximate Query Processing Engines with Machine Learning Models. In SIGMOD '19.	Jiaheng

Academic Integrity

All assignments (pre-lecture questions, assignments and presentation slides) must be prepared and written independently and reflect the student's own opinions. Simply paraphrasing other students' work is considered as plagiarism and we take the recommended actions for any discovered academic integrity violation per Departmental and University policies, including receiving an F grade or other appropriate penalties depending on the severity of the violation. While discussion among students prior to and after lectures are allowed, it is your responsibility to ensure that your submission is not substantially similar to any other student's submission. Note that it is generally acceptable to use part or all of the presentation slides found on conference website or the author's website, as long as there are proper citations and acknowledgments.