Skip to content

Details

Join us on the 3rd of December for the second BEADS tech meetup on realtime analytics, co-organized with SAP, featuring speakers from ClickHouse, SAP, and HPI!

Program:

Doors open at 18.00 and talks start at 18.30:

  • Robert Schulze (ClickHouse) — Three Clever Ways How Analytics Databases Cheat. Tip: They all have to do with pruning.
  • Thomas Bodner (SAP) — Real-time Analytics for Enterprise Applications with SAP HANA.
  • Martin Boissier (HPI) — SmartRabbit: An Interactive Query Processor

Networking and drinks to follow.

Details on the Talks and Speakers:

Three clever ways how analytics databases cheat. Tip: They all have to do with pruning. Analytics databases like ClickHouse routinely filter, aggregate and join terabyte-sized tables. This is only possible through massively parallel data processing (SIMD, multi-threading, distributed query execution), or by "cheating" - that is, by not processing data altogether. In this presentation, we demonstrate three ways how ClickHouse uses pruning to avoid reading data as much as possible.

Robert Schulze is an engineering manager at ClickHouse, where he works on query performance and optimization. Before joining ClickHouse, he was an engineer in the SAP HANA development team. He holds a diploma in computer science from TU Dresden.

Real-time Analytics for Enterprise Applications with SAP HANA. Enterprise applications need analytical queries to return quickly and reflect the current state of the business. Meeting these requirements calls for efficient queries over complex business models, cloud resources that adapt to changing demand, and economical access to growing datasets. This talk examines how these requirements shape query optimization, execution, and storage management in SAP HANA. We show how reusable business views can turn a simple analytical question into a query with dozens of joins, and how the optimizer simplifies the resulting plans. We then discuss how elastic compute provides additional capacity for analytical workloads and how combining in-memory processing with disk-based columnar storage reduces memory requirements and costs for large datasets. Together, these examples illustrate how application and database co-design makes real-time analytics practical for enterprise applications.

Thomas Bodner is a software architect in SAP’s Berlin office, where he helps to make SAP HANA scale more elastically and automatically in the cloud. His doctoral research at Hasso Plattner Institute focused on data processing on serverless cloud infrastructure. He holds a Master’s degree in computer science from TU Berlin. Thomas got hooked on databases early on during an undergraduate internship at IBM Almaden, as part of a program originally established by Volker Markl.

SmartRabbit: An Interactive Query Processor. Traditional relational database systems optimize analytical queries to minimize their end-to-end latency. The resulting optimal plans are usually blocking, forcing users to wait until full query completion before seeing any results. This execution model precludes interactivity, i.e., users cannot observe partial results or gain early insights for long-running queries. Query optimizers rarely choose plans that promote interactivity, since such plans either incur prohibitively large latencies or involve operators for which interactive alternatives are often infeasible. We present a novel interactive query processor named SmartRabbit that promotes interactivity of answers while matching the end-to-end latency of blocking execution plans. We achieve this by first designing a plan optimized for interactivity for a given query, and then simultaneously executing this plan alongside a traditional blocking plan. The two executions are carefully synchronized to maintain the correct order of answers and prevent duplicates. We implement SmartRabbit in a scalable, open-source database system and show that SmartRabbit consistently delivers early and continuous results across various analytical benchmarks, data scales, and levels of parallelism, with only marginal latency overhead compared to traditional blocking execution.

Martin Boissier did his PhD at the chair of Prof. Hasso Plattner, working on optimizing compression schemes for analytical in-memory database systems. From 2023 to 2024, he visited UC Irvine and joined the database group of Prof. Sharad Mehrotra. Currently, Martin is a PostDoc at the chair of Prof. Tilmann Rabl’s data engineering systems group. He is heading the Hyrise database project.

Related topics

Events in Berlin, DE
Artificial Intelligence
Big Data
Data Analytics
Data Science
Database Development

You may also like