Details
-
New Feature
-
Status: Open
-
Critical
-
Resolution: Unresolved
Description
Synopsis
The current Apache Wayang (Incubating) uses a cost model to select the right set of platforms while optimize query plans. Often, the initial cost model could be ineffective after some time, and a calibration of the cost model is required again. The goal is to create a pipeline that starts a ML pipeline that starts the calibration of the cost model automatically and uses the logs of the previous query executions to get refine the cost model so that it follows the workload that interacts with the Apache Wayang (Incubating) environment.
Benefits to Community
The benefits for the community will have an AI pipeline for automatic, dynamic cost model calibration in query optimizers; We will use Apache Wayang(Incubating) as our playground. As a result, the experience of the users of Apache Wayang(Incubating) will improve by helping them to automatically tune their cost models and adatp to the current query workload.
Deliverables
The delivery expected is an adaptation for the paper "Zero-Shot Cost Models for Out-of-the-box Learned Cost Prediction"[1], where the authors assume an ML-Cost-Model. Still, in this case, the idea needs modifications to run in the current setup of Apache Wayang(Incubating).
The step expected are the following:
- Understand the paper [1]
- Get into the cost model of Apache Wayang
- Discuss and design the process for the dynamic cost-model
- Implement the feature of dynamic cost-model
Related Work
[1] [Zero-Shot Cost Models for Out-of-the-box Learned Cost Prediction](https://arxiv.org/pdf/2201.00561.pdf)
[2] [RHEEMix in the data jungle: a cost-based optimizer for cross-platform systems](https://wayang.apache.org/assets/pdf/paper/journal_vldb.pdf)
Biographical Information of possible mentor{}
Bertty Contreras-Rojas is a Senior Software Engineer at Databloom Inc. He is one of the PPMC of Apache Wayang(Incubating). He has many years of experience developing intensive processing data systems for several industries, such as banking systems. He was a research engineer at the Qatar Computing Research Institute, where he was responsible for developing the declarative query engine for Rheem and adding new underlying platforms to Rheem.
Rodrigo Pardo-Meza is a Senior Software Engineer at Databloom Inc. He is one of the PPMC of Apache Wayang(Incubating). He has many years of experience developing applications that support Big Data processing, with experience implementing ETL processes over distributed systems to optimize inventories in supply chains. He was a research engineer at the Qatar Computing Research Institute, where he specialized in human interface interaction with big data analytics. During this time, he co-develop an ML-based cross-platform query optimizer.
Jorge Quiané is the head of the Big Data Systems research group at the Berlin Institute for the Foundations of Learning and Data (BIFOLD) and a Principal Researcher at DIMA (TU Berlin). He also acts as the Scientific Coordinator of the IAM group at the German Research Center for ArtificialIntelligence (DFKI). His current research is in the broad area of big data: mainly in federated data analytics, scalable data infrastructures, and distributed query processing. He has published numerous research papers on data management and novel system architectures. He has recently been honoured with the 2022 ACM SIGMOD Research Highlight Award and the Best Paper Award at ICDE 2021 for his work on “EfficientControl Flow in Dataflow Systems”. He holds five patents in core database areas and on machine learning. Earlier in his career, he was a Senior Scientist at the Qatar Computing Research Institute (QCRI) and a Postdoctoral Researcher at Saarland University. He obtained his PhD in computer science from INRIA (Nantes University).
Name and Contact Information
Name: Bertty Contreras-Rojas
email: bertty (at) apache.org
community: dev (at) wayang.apache.org
website: https://wayang.apache.org
