Web Search for a Planet: The Google Cluster Architecture

By Luiz André Barroso, Joseph Dean, Urs Hölzle. IEEE Micro, 2003

1,122 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/c1a0e40c06392f263bb2614d0d5ed290c3b3602c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Information Systems

Citations in the corpus

Is searching full text more effective than searching abstracts? (2009, background): “Even though outsiders are occasionally offered glimpses into their design [46,47 MapReduce draws inspiration from higher-order functions in functional programming and builds on the observation that many information processing tasks have the same basic structure: a computation is applied over a…”
Making Scheduling "Cool": Temperature-Aware Workload Placement in Data Centers (2005, background): “…1U servers consume over 350 Watts at peak utilization, releasing much of this energy as heat; a standard 42U rack of such servers consumes over 8 kW. Barroso et al estimate that the power density of the Google data center is three to ten times that of typical commercial data centers [10].”
Characterizing cloud computing hardware reliability (2010, background): “Characterizing Cloud Computing Hardware Reliability Kashi Venkatesh Vishwanath and Nachiappan Nagappan Microsoft Research One Microsoft Way, Redmond WA 98052 {kashi.vishwanath,nachin}@microsoft.com ABSTRACT Modern day datacenters host hundreds of thousands of servers that coordinate tasks in order”
Total recall II: Query expansion revisited (2011, methodology): “The search engines for particular object search have been inspired by widely used text search engines [2, 4].”
Disaggregated memory for expansion and sharing in blade servers (2009, background): “…we are likely to see an increased number of virtual machines (VMs) per core (VMware quotes 2-4X memory requirements from VM consolidation every generation [2]), and increased memory footprint per VM (e.g., the footprint of Microsoft® Windows® has been growing faster than Moore’s Law [3]).”
Deeper Inside PageRank (2004, background): “The 1998 paper by Brin and Page [26] and more recent papers by Google engineers [13, 56] provide detailed discussions of the many storage schemes used by the Google search engine for all parts of its information retrieval system.”
Evaluating MapReduce for Multi-core and Multiprocessor Systems (2007, background): “Google's runtime implementation targets large clusters of Linux PCs connected through Ethernet switches [3].”
Remote sensing big data computing: Challenges and opportunities (2015, background): “Google [44] has build a large cluster system consisting of nearly 1500 ordinary personal computers.”
A Survey on Distributed Machine Learning (2019, background): “This strategy has proven more aordable compared to using more expensive specialized hardware, as long as sucient fault tolerance is built into the software, a concept that Google has pioneered [16] and that has increasingly found traction in the industry.”
Jupiter Rising: A Decade of Clos Topologies and Centralized Control in Google's Datacenter Network (2015, background): “A combination of remote storage access [7,14], large-scale data processing [10,18], and interactive web services [4] drive our bandwidth demands.”
Efficient indexing for large scale visual search (2009, background): “In contrast, it is reported that a commercial web search engine can index more than 5 million web pages by a 32-bit machine with the same data structure, and answer a query in a few milliseconds [2].”
The RAMCloud Storage System (2015, methodology): “For example, Google keeps its entire Web search index in DRAM [Barroso et al. 2003], Facebook offloads its database servers by caching tens of terabytes of data in DRAM with memcached [2011], and Bigtable allows entire column families to be loaded into memory [Chang et al. 2008].”
Collaborative Filtering: A Machine Learning Perspective (2004, neutral): “Suppose that we were given computational resources equivalent to those currently used by Google: 10(4) CPU’s running at 500MHz each capable of performing 10(8) additions and subtractions per second [1].”
Massively multi-robot simulation in stage (2008, background): “The greatest opportunity for massive scaling comes with cluster computing, in which many hosts can be pooled into a distributed system (e.g., Google's huge systems; see Barroso et al. 2003).”