Hardware and Software Co-design for HTAP Workloads
Event details
| Date | 26.06.2018 |
| Hour | 14:30 › 16:30 |
| Speaker | Zilu Tian |
| Location | |
| Category | Conferences - Seminars |
EDIC candidacy exam
Exam president: Prof. Christoph Koch
Thesis advisor: Prof. Babak Falsafi
Co-examiner: Prof. Paolo Ienne
Abstract
Real-time analytics has gained more visibility in recent years. Instead of relying on historical data for decision making, some business intelligence applications demand analysis on fresh data to gain critical insights, highlighting the rise of mixed Online Transaction Processing (OLTP) and Online Analytical Processing (OLAP) workloads, which are coined Hybrid Transactional/Analytical Processing (HTAP) workloads. The OLTP workload is characterized by short-lived transactions, where data is stored in a row-oriented database. The OLAP workload consists of long-running, scan-heavy, ad-hoc queries, where data is stored in a column-oriented layout. The OLTP and OLAP workloads have conflicting needs for latency and throughput. No single physical layout fits the hybrid nature of HTAP workloads. In this project, we explore hardware and software co-design to accelerate HTAP workloads.
Background papers
Gather-Scatter DRAM:In-DRAM Address Translation to Improve the Spatial Locality of Non-unit Strided Access, by Vivek Seshadr et al.
Column-Stores vs Row-Stores: How Different Are They Really? by Daniel J. Abadi et al.
Bridging the Archipelago between Row-Stores and Column-Stores for Hybrid Workloads, by Joy Arulraj et al.
Exam president: Prof. Christoph Koch
Thesis advisor: Prof. Babak Falsafi
Co-examiner: Prof. Paolo Ienne
Abstract
Real-time analytics has gained more visibility in recent years. Instead of relying on historical data for decision making, some business intelligence applications demand analysis on fresh data to gain critical insights, highlighting the rise of mixed Online Transaction Processing (OLTP) and Online Analytical Processing (OLAP) workloads, which are coined Hybrid Transactional/Analytical Processing (HTAP) workloads. The OLTP workload is characterized by short-lived transactions, where data is stored in a row-oriented database. The OLAP workload consists of long-running, scan-heavy, ad-hoc queries, where data is stored in a column-oriented layout. The OLTP and OLAP workloads have conflicting needs for latency and throughput. No single physical layout fits the hybrid nature of HTAP workloads. In this project, we explore hardware and software co-design to accelerate HTAP workloads.
Background papers
Gather-Scatter DRAM:In-DRAM Address Translation to Improve the Spatial Locality of Non-unit Strided Access, by Vivek Seshadr et al.
Column-Stores vs Row-Stores: How Different Are They Really? by Daniel J. Abadi et al.
Bridging the Archipelago between Row-Stores and Column-Stores for Hybrid Workloads, by Joy Arulraj et al.
Practical information
- General public
- Free
Contact
- EDIC - [email protected]