Energy-efficient cluster computing with FAWN: workloads and implications

@inproceedings{Vasudevan2010EnergyefficientCC,
  title={Energy-efficient cluster computing with FAWN: workloads and implications},
  author={Vijay Vasudevan and David G. Andersen and Michael Kaminsky and Lawrence Tan and Jason Franklin and Iulian Moraru},
  booktitle={e-Energy},
  year={2010}
}
This paper presents the architecture and motivation for a cluster-based, many-core computing architecture for energy-efficient, data-intensive computing. FAWN, a Fast Array of Wimpy Nodes, consists of a large number of slower but efficient nodes coupled with low-power storage. We present the computing trends that motivate a FAWN-like approach, for CPU, memory, and storage. We follow with a set of microbenchmarks to explore under what workloads these "wimpy nodes" perform well (or perform poorly… 

Figures and Tables from this paper

Improving Datacenter Energy Efficiency Using a Fast Array of Wimpy Nodes
TLDR
This thesis proposes to help reduce the energy consumed by large-scale computing by using a FAWN: A Fast Array of Wimpy Nodes, and investigates the applicability of FAWN to data-intensive workloads.
The search for energy-efficient building blocks for the data center
TLDR
This paper conducts a survey of several small clusters of machines in search of the most energy-efficient data center building block targeting data-intensive computing, and builds five-node homogeneous clusters of each type and runs Dryad, a distributed execution engine, with a collection of data- intensive workloads to measure the energy consumption per task.
Energy-efficient Data-intensive Computing with a Fast Array of Wimpy Nodes
TLDR
Using FAWN as an early example, it is demonstrated that pervasive use of vector interfaces throughout distributed storage systems can improve throughput by an order of magnitude and eliminate the redundant work found in many data-intensive workloads.
Enabling efficiency in data center systems
TLDR
This dissertation presents the storage configuration compiler that combines an application model and workload characterization to identify low-cost hardware configurations and presents Stout, a software layer that enables more efficient use of shared storage.
FAWN-DLi : A Data Library for a Fast Array of Wimpy Nodes
TLDR
This project extends one existing lowpower system, FAWN-KV, to support more sophisticated sever-side data operations and improves performance by up to a factor of three and outperformed the basic system in all tests.
An Experimental Evaluation of Datacenter Workloads On Low-Power Embedded Micro Servers
TLDR
This paper presents a comprehensive evaluation of an ultra-low power cluster, built upon the Intel Edison based micro servers, which shows that the Edison cluster achieves up to 3.5x improvement on work-done-per-joule for web service applications and data-intensive MapReduce jobs.
Flashing in the Cloud
  • Jalil Boukhobza
  • Computer Science
    Data Intensive Storage Services for Cloud Environments
  • 2013
TLDR
The purpose of this chapter is to introduce flash memory storage systems by focusing on their specific architectures and algorithms, and finally their integration into servers and data centers.
SWEEP: evaluating computer system energy efficiency using synthetic workloads
TLDR
SWEEP, Synthetic Workloads for Energy Efficiency and Performance evaluation, a framework for generating synthetic workloads with specific behavioral characteristics, is proposed and the Energy-Delay Diagram (EDD) is presented, a novel method for visualizing energy efficiency.
CHERUB: power consumption aware cluster resource management
  • Simon Kiertscher, J. Zinke, B. Schnor
  • Computer Science
    2010 IEEE/ACM Int'l Conference on Green Computing and Communications & Int'l Conference on Cyber, Physical and Social Computing
  • 2010
TLDR
An evaluation of ACPI energy saving modes, and the design and implementation of an energy saving daemon for clusters called cherub are deduced, which shows that cherub’s scheduling algorithm works well, and it will save energy, if possible, and avoids state-flapping.
Wimpy node clusters: what about non-wimpy workloads?
TLDR
Results show that in most cases, computationally complex queries exhibit disproportionate scaleup characteristics which potentially makes scale-out with low-end nodes an expensive and lower performance solution.
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 29 REFERENCES
FAWN: a fast array of wimpy nodes
TLDR
The key contributions of this paper are the principles of the FAWN architecture and the design and implementation of FAWN-KV--a consistent, replicated, highly available, and high-performance key-value storage system built on a FAWN prototype.
Gordon: using flash memory to build fast, power-efficient clusters for data-intensive applications
TLDR
The paper presents an exhaustive analysis of the design space of Gordon systems, focusing on the trade-offs between power, energy, and performance that Gordon must make, and describes a novel flash translation layer tailored to data intensive workloads and large flash storage arrays.
Low-power amdahl-balanced blades for data intensive computing
TLDR
This paper proposes an alternative architecture comprising large number of so-called Amdahl blades that combine energy-efficient CPUs with solid state disks to increase sequential read I/O throughput by an order of magnitude while keeping power consumption constant.
Active Disks for Large-Scale Data Processing
TLDR
This work proposes using an active disk storage device that combines on-drive processing and memory with software downloadability to allow disks to execute application-level functions directly at the device.
Low Power Amdahl Blades for Data Intensive Computing
TLDR
The emergence of Solid State Disk (SSD) technology, poses the challenge of building a credible equivalent to a GrayWulf system with a similar IO performance, but with considerably lower power consumption, than an SSD-based system, while maintaining the same cost and power consumption.
Web Search Using Small Cores: Quantifying the Price of Efficiency
TLDR
This work evaluates search on server and mobile-class architectures using Xeon and Atom processors, quantifying search efficiency at the microarchitecture and system-level and prompts us toward re-thinking small core designs for a new breed of data center workloads in order to continue reaping the benefits of small-core power efficiency.
Hibernator: helping disk arrays sleep through the winter
TLDR
This paper describes the Hibernator design, and presents evaluations of it using both trace-driven simulations and a hybrid system comprised of a real database server (IBM DB2) and an emulated storage server with multi-speed disks.
JouleSort: a balanced energy-efficiency benchmark
TLDR
This work proposes and motivate JouleSort, an external sort benchmark, for evaluating the energy efficiency of a wide range of computer systems from clusters to handhelds, and demonstrates a Joule sort system that is over 3.5x as energy-efficient as last year's estimated winner.
Understanding and Designing New Server Architectures for Emerging Warehouse-Computing Environments
TLDR
A new solution that incorporates volume non-server-class components in novel packaging solutions, with memory sharing and flash-based disk caching, has promise, with a 2X improvement on average in performance-per-dollar for the benchmark suite.
Evaluation of Existing Architectures in IRAM Systems
TLDR
This work examined both execution time analyses of existing microprocessors and system simulation of hypothetical processors to determine whether existing microarchitectures can tap the potential performance advantages of IRAM systems.
...
1
2
3
...