PDF
You should note that the tile of this book is "The Datacenter as A Computer", not "ComputerS in a datacenter". The authors introduce a new concept of Warehouse-Scale Computers (WSCs), saying that a data center should be seen as a computer itself (having organic solidarity between its own physical computers) not simply a set of independent computers.
This article was written by Google insiders, so many explanations are based on the case of Google (we can sneak a look into how Google does). But they also fairly mention about their competitors, such as Microsoft, Yahoo, and Amazon.
Chapter 1 - Introduction
It begins with some basic concepts of datacenter computing (what it is and how it is different from "traditional" datacenters). A brief architectural overview of WSCs is given, mostly from the perspective of hardware. Because WSCs highly depend on the economy of scale, the book focuses on how to be the most effective given constraints of commodity hardware. The authors leave many design issues, which are addressed in the rest of the book.
Chapter 2 - Workloads and Software Infrastructure
To simplify the inherent complexity of WSCs, the authors break its software infrastructure into three layers: i) platform-level software (for each machine), ii) cluster-level infrastructure (abstraction layer for common tasks), and iii) application-level software (application-specific). This hierarchy of software infrastructure is designed to handle the differences of workloads, requirements, and constraints from traditional single-server applications.
This chapter introduces some useful techniques to build such infrastructure. The authors also gives two examples (online and offline applications) to show what should be considered in the design process. They also address the monitoring issues of the system since it is not easy to figure out the status and problems in such a huge system.
You should note that the tile of this book is "The Datacenter as A Computer", not "ComputerS in a datacenter". The authors introduce a new concept of Warehouse-Scale Computers (WSCs), saying that a data center should be seen as a computer itself (having organic solidarity between its own physical computers) not simply a set of independent computers.
This article was written by Google insiders, so many explanations are based on the case of Google (we can sneak a look into how Google does). But they also fairly mention about their competitors, such as Microsoft, Yahoo, and Amazon.
Chapter 1 - Introduction
It begins with some basic concepts of datacenter computing (what it is and how it is different from "traditional" datacenters). A brief architectural overview of WSCs is given, mostly from the perspective of hardware. Because WSCs highly depend on the economy of scale, the book focuses on how to be the most effective given constraints of commodity hardware. The authors leave many design issues, which are addressed in the rest of the book.
Chapter 2 - Workloads and Software Infrastructure
To simplify the inherent complexity of WSCs, the authors break its software infrastructure into three layers: i) platform-level software (for each machine), ii) cluster-level infrastructure (abstraction layer for common tasks), and iii) application-level software (application-specific). This hierarchy of software infrastructure is designed to handle the differences of workloads, requirements, and constraints from traditional single-server applications.
This chapter introduces some useful techniques to build such infrastructure. The authors also gives two examples (online and offline applications) to show what should be considered in the design process. They also address the monitoring issues of the system since it is not easy to figure out the status and problems in such a huge system.
No comments:
Post a Comment