报告题目：Approximate Query Processing via Sampling

Abstract:

Join operator is one of the most important one in relational algebra one. Yet, it is often very time-consuming to compute in full on a large database. On the other hand, it is often the case that a user

only needs a fairly accurate approximate query result for an aggregation query instead of an exact answer.

Towards that, we have investigated how we can efficiently retrieve samples directly from the join results

without computing them in full and how to utilize them to answer approximate queries. In this talk,

I will present two algorithms that apply to different application scenarios, namely the Wander Join

algorithm which produces non-uniform but independent samples for approximate SPJA queries,

and a general random sampling framework that produces uniform and independent samples for other

types of complex analytical tasks. Experimental evaluation in real database systems show the algorithms

outperforms the state-of-the-art by orders of magnitude on the TPC-H benchmark dataset.

Bio:

Zhuoyue Zhao is a fourthyear Ph.D. student at University of Utah and his advisor is Prof. Feifei Li.

He is interested in query processing and query optimization in OLAP databases and his main focus

is on approximate query processing and sampling from complex queries. Zhuoyue received the

best paper award in SIGMOD'16 and received the Google PhD Fellowship in 2019.

He received B.S.E degree from Shanghai Jiao Tong University.