Data Processing On Modern Hardware
The lectures for this course are pre-recorded and available on the Moodle course web-page.
The tutorials are going to be held via video web-conference.
This course highlights some of the implications that current hardware trends have on database processing. Advances such as deep cache hierarchies or the
use of hardware accelerators had a major impact on how we design and implement data processing algorithms and data structures. This lecture will show how
carefully laying out data in memory and good algorithm design can increase the effectiveness of hardware caches; how we can speed up database operations
by parallelizing on modern CPUs; how to achieve efficient synchronization for data structures; and how to leverage specialized instructions/accelerators
(e.g., SIMD) for data processing. We are also going to take a look at offloading computation to programmable hardware devices (FPGAs), and see how we can
benefit from novel network and storage technologies (RDMA and NVRAM).
More specifically, the topics we are going to cover are the following topics:
- Writing efficient code for the memory hierarchy.
- Parallelizing data-intensive tasks on multi-core CPUs.
- Using efficient synchronization of data structures.
- Leveraging modern hardware features and technologies for compute (e.g., SIMD processing, accelerators), network (e.g., RDMA), and storage (e.g., NVRAM).
- 5 ECTS
- SWS 2V + 2Ü
- Lectures are held in English
- Lectures uploaded on Tuesdays afternoon to Moodle.The official lecture slot is on Wednesdays between 9:00 am - 10:30 am
- The tutorial is held via BBB on Wednesdays from 10:30 am - 12:00 pm
- For the Gitlab repository see Moodle.
- For the Mattermost channel see Moodle.
- The assessment is in the form of a virtual oral exam. Bonus will be given to students who do the exercise assignments and project work.
The course is aimed at Master-level students who have solid systems programming experience in C/C++ and have already taken the following (or similar) courses:
- Introduction to Databases
- Introduction to Computer Architecture
Slides will be regularly uploaded shortly before each lecture.
- Lecture 1: Introduction and Hardware Trends (videos)
- Lecture 2: Cache Awareness (videos)
- Lecture 3: Cache Awareness for Query Execution Models (1st videos, 2nd videos)
- Lecture 4: Memory-intensive operations (1st videos, 2nd videos)
- Lecture 5: Instruction execution (1st videos, 2nd videos)
- Lecture 6: Data-Level Parallelism (1st videos, 2nd videos)
- Lecture 7: Multicore Parallelism and Synchronization (videos)
- Lecture 8: Multicore CPUs: NUMA, interference, and isolation (videos)
- Lecture 10: Rack-scale data processing (videos)
- Assignment 1: Cache Awareness
- Assignment 2: Query Execution Models
- Assignment 3: Hardware Optimized Hash Joins
- Assignment 4: SIMD Vectorization
- Assignment 5: Synchronization
- Assignment 6: Task-Level Parallelism
This is not a standard course (i.e., there is no real textbook). Most material is taken out of research papes, which will be referenced in the slides. However, the following list can be useful either as background or complementary reading.
- "Computer Architecture: A Quantitative Approach" (6th edition) by Hennessy and Petterson.
- "Computer Systems: A Programmer's Perspective" (3rd edition) by Bryant and O'Hallaron
- Intel's Software Developer Manuals
- Intel's Top-Down Microarchitectural Analysis Method (TMAM) and the Roofline model
- Agner Fog's Software optimization resources
- Ulrich Drepper's What Every Programmer Should Know About Memory