Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
The datacenter as a computer : designing warehouse-scale machines (3rd ed.)
Barroso L., Hölzle U., Ranganathan P., Szefer J., Morgan&Claypool Publishers, San Rafael, CA, 2019. 207 pp. Type: Book (978-1-681734-33-0)
Date Reviewed: Mar 29 2019

As machine learning becomes more popular and more responsible for managing our everyday lives, using it to manage data centers, to control the mammoth tasks engineers have to deal with, is only a natural evolution.

This book is a series of lectures in computer architecture, with each chapter devoted to a particular aspect of the data center and the computational challenges in managing it. Written by notable authors in the field, the chapters describe the cost efficiency versus compute tradeoff, managing multiple machine learning workloads, latency, energy efficiency, and even how to manage cooling racks to keep a machine active and working over time.

The book concentrates on Google data centers, the software stack, and tensor processing units (TPUs), and might have benefited from exposing some of the related data on efficiency and performance. However, it does discuss how computational complexity has risen over the years, and there is a need to come up with innovative ways in engineering to manage this.

There are a few intersecting sections in the book, particularly on software-defined infrastructure and how it can make life easier for the engineer to monitor performance metrics and deduce which part of the data center needs closer attention. The book mostly concentrates on hardware and software stacks, with little on the network optimizations within the data center; however, there is already a large amount of literature devoted to this.

Another interesting discussion is on diagnosing faults, classifying these, and designing recovery tactics to prevent faults from disrupting workflow computations. The authors present new ways to document faults based on the damage they can do, and decide the time to react to these.

The book is an interesting read overall, particularly for intermediary-level engineers who are already managing data centers but want to do it more efficiently, for example, by learning about cost/time savings from Google. Getting the most computation for your buck is a challenge we face today, and this book attempts to solve some of these issues with an in-depth technical discussion that will be useful to all readers.

More reviews about this item: Goodreads

Reviewer:  Mariam Kiran Review #: CR146501 (1906-0207)
Bookmark and Share
  Featured Reviewer  
 
General (C.2.0 )
 
 
Cloud Computing (C.2.4 ... )
 
 
Distributed Artificial Intelligence (I.2.11 )
 
Would you recommend this review?
yes
no
Other reviews under "General": Date
Comparative evaluation of networks and protocols
Pooch U., John Wiley & Sons, Inc., New York, NY, 1984. Type: Book (9780471262336)
Aug 1 1985
Data communications, networks, and systems
Bartee T., Sams, Indianapolis, IN, 1985. Type: Book (9780672222351)
May 1 1987
Introduction to data communications and computer networks
Halsall F. (ed), Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1985. Type: Book (9789780201145472)
May 1 1986
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy