Tutorial on dist-gem5 at ISCA 2017
Sunday, June 25, 9:00 to 12:30
44th International Symposium on Computer Architecture, June 24-28, 2017, Toronto, ON, CanadaContents
List of organisers/presenters
- Nam Sung Kim, University of Illinois, Urbana-Champaign
- Mohammad Alian, University of Illinois, Urbana-Champaign
- Nikos Nikoleris, ARM Ltd.
- Radhika Jagtap, ARM Ltd.
- Gabor Dozsa, ARM Ltd.
- Stephan Diestelhorst, ARM Ltd.
Abstract
The single-thread performance improvement of processors has been sluggish for the past decade as Dennard’s scaling is approaching its fundamental physical limit. Thus, the importance of efficiently running applications on a parallel/distributed computer system has continued to increase and diverse applications based on parallel/distributed computing models such as MapReduce and MPI have thrived.
In a parallel/distributed computing system, the complex interplay amongst processor, node, and network architectures strongly affects the performance and power efficiency. In particular, we observe that all the hardware and software aspects of the network, which encompasses interface technology, switch/router capability, link bandwidth, topology, traffic patterns, and protocols, significantly impact the processor and node activities. Therefore, to maximize performance and power efficiency, it is critical to develop various optimization strategies cutting across processor, node, and network architectures, as well as their software stacks, necessitating full-system simulation. However, our community lacks a proper research infrastructure to study the interplay of these subsystems. Facing such a challenge, we have released a gem5-based simulation infrastructure dubbed dist-gem5 to support full-system simulation of a parallel/distributed computer system using multiple simulation host. This tutorial will cover an introduction to dist-gem5 including relevant background knowledge.
Objectives
More specifically, the tutorial will provide the following.
- Introduction of parallel/distributed system architecture.
- Details of enhanced gem5 components to enable simulation of a parallel/distributed computer system.
- Network interface and switch models to connect multiple simulated nodes (as shown in the Figure).
- Synchronization amongst multiple simulated nodes running across multiple simulation hosts.
- Simulating a region of interest of a given benchmark using check-point creation/restoration enhanced for simulating multiple simulated nodes using multiple simulation hosts.
- Examples of modeling parallel/distributed computer systems using a few network topologies.
| 09:00 – 10:00 | Introduction (60 min) |
| 10:00 – 10:15 | Break (15 min) |
| 10:15 – 11:15 | dist-gem5 deep dive (60 min) |
| 11:15 – 11:30 | Break (15 min) |
| 11:30 – 12:00 | dist-gem5 examples (30 min) |
Slides
- The slides from the tutorial can be downloaded here.
Publications
- Mohammad Alian, Gabor Dozsa, Umur Darbaz, Stephan Diestelhorst, Daehoon Kim, and Nam Sung Kim. “dist-gem5: Distributed Simulation of Computer Clusters”, IEEE International Symposium on Performance Analysis of Systems (ISPASS), April 2017 (Nominated for the Best Paper Award)
- Mohammad Alian, Daehoon Kim, and Nam Sung Kim. “pd-gem5: Simulation Infrastructure for Parallel/Distributed Computer Systems”, IEEE Computer Architecture Letters (CAL), Jan 2016 paper
Pre-requisites
- Basic knowledge of computer architecture
- No prior experience with simulators is required
