Google data centers: Difference between revisions

From Wikipedia, the free encyclopedia
Browse history interactively
← Previous editNext edit →
Content deleted Content added
VisualWikitext
Undid revision 423149835 by Alpha7248 (talk) read the whole sentence
m →‎Software: improved wording
Line 69: Line 69:


Google has developed several abstractions which it uses for storing most of its data:<ref name="quinlan">http://www.eweekeurope.co.uk/news/news-it-infrastructure/google-developing-caffeine-storage-system-1620</ref>
Google has developed several abstractions which it uses for storing most of its data:<ref name="quinlan">http://www.eweekeurope.co.uk/news/news-it-infrastructure/google-developing-caffeine-storage-system-1620</ref>
* [[Protocol buffers]] &mdash; "Google's lingua franca for data"<ref>http://code.google.com/apis/protocolbuffers/docs/overview.html</ref>, a widely used within the company binary serialization format.
* [[Protocol buffers]] &mdash; "Google's lingua franca for data"<ref>http://code.google.com/apis/protocolbuffers/docs/overview.html</ref>, a binary serialization format which is widely used within the company.
* SSTable (Sorted Strings Table) &mdash; a persistent, ordered, immutable map from keys to values, where both keys and values are arbitrary byte strings. It is also used as one the building blocks of BigTable.<ref>http://labs.google.com/papers/bigtable-osdi06.pdf</ref>
* SSTable (Sorted Strings Table) &mdash; a persistent, ordered, immutable map from keys to values, where both keys and values are arbitrary byte strings. It is also used as one the building blocks of BigTable.<ref>http://labs.google.com/papers/bigtable-osdi06.pdf</ref>
* RecordIO &mdash; a sequence of variable sized records.<ref>http://www.windley.com/archives/2008/06/velocity_08_storage_at_scale.shtml</ref><ref name="quinlan" /><ref>http://groups.google.com/group/protobuf/browse_thread/thread/ee27572aef9da70a</ref>
* RecordIO &mdash; a sequence of variable sized records.<ref>http://www.windley.com/archives/2008/06/velocity_08_storage_at_scale.shtml</ref><ref name="quinlan" /><ref>http://groups.google.com/group/protobuf/browse_thread/thread/ee27572aef9da70a</ref>

Revision as of 06:11, 11 April 2011

Google's first production server rack, circa 1999

Google requires large computational resources in order to provide their services. This article describes the technological infrastructure behind Google's websites, as presented in the company's public announcements.

Hardware

Original hardware

The original hardware (circa 1998) that was used by Google when it was located at Stanford University included:[1]

  • Sun Ultra II with dual 200 MHz processors, and 256 MB of RAM. This was the main machine for the original Backrub system.
  • 2 × 300 MHz Dual Pentium II Servers donated by Intel, they included 512 MB of RAM and 9 × 9 GB hard drives between the two. It was on these that the main search ran.
  • F50 IBM RS/6000 donated by IBM, included 4 processors, 512 MB of memory and 8 × 9 GB hard drives.
  • Two additional boxes included 3 × 9 GB hard drives and 6 x 4 GB hard drives respectively (the original storage for Backrub). These were attached to the Sun Ultra II.
  • IBM disk expansion box with another 8 × 9 GB hard drives donated by IBM.
  • Homemade disk box which contained 10 × 9 GB SCSI hard drives.

Current hardware

Servers are commodity-class x86 PCs running customized versions of Linux. The goal is to purchase CPU generations that offer the best performance per dollar, not absolute performance, how this is measured is unclear but is likely to incorporate running costs of the entire server and CPU power consumption could be significant factor.[2] Servers as of 2009 consisted of a custom made open top server containing two processors (each with an unknown number of cores or interconnected processing units) a considerable amount of RAM spread over 8 DIMM slots housing double height DIMMS and two SATA hard drives connected through a standard ATX sized power supply. Each server has a novel 12 volt battery to reduce costs and improve power efficiency [3]

Estimates of the power required for over 450,000 servers range upwards of 20 megawatts, which cost on the order of US$2 million per month in electricity charges. The combined processing power of these servers might reach from 20 to 100 petaflops.[4]

Specifications:

  • In 2002; upwards of 15,000 servers[5] ranging from 533 MHz Intel Celeron to dual 1.4 GHz Intel Pentium III (as of 2003).
  • One or more 80 GB hard disks per server (2003)
  • 2–4 GB of memory per machine (2004)
  • A 2005 estimate by Paul Strassmann has 200,000 servers,[6] while unspecified sources claimed this number to be upwards of 450,000 in 2006.[7]
  • ~ 16 GB RAM, 2 TB disk space per machine (2009)[8]

The exact size and whereabouts of the data centers Google uses are unknown, and official figures remain intentionally vague. A very old estimate (from 2000 while Google was in its infancy and had one product), Google's server farm consisted of 6,000 processors, 12,000 common IDE disks (2 per machine, and one processor per machine), at four sites: two in Silicon Valley, California and one in Virginia.[9] Each site had an OC-48 (2488 Mbit/s) internet connection and an OC-12 (622 Mbit/s) connection to other Google sites. The connections are eventually routed down to 4 × 1 Gbit/s lines connecting up to 64 racks, each rack holding 80 machines and two Ethernet switches. [citation needed]

Hardware details considered sensitive

In a 2008 book,[10] reporter Randall Stross wrote: "Google's executives have gone to extraordinary lengths to keep the company's hardware hidden from view. The facilities are not open to tours, not even to members of the press." He wrote this based on interviews with staff members and his experience of visiting the company.

Network topology

When a client computer attempts to connect to Google, several DNS servers resolve www.google.com into multiple IP addresses via Round Robin policy. Furthermore, this acts as the first level of load balancing and directs the client to different Google clusters. A Google cluster has thousands of servers and once the client has connected to the server additional load balancing is done to send the queries to the least loaded web server. This makes Google one of the largest and most complex content delivery networks.[11]

Racks are custom-made and contain 40 to 80 servers (20 to 40 1U servers on either side), while new servers are 2U Rackmount systems.[5] Each rack has a switch. Servers are connected via a 100 Mbit/s Ethernet link to the local switch. Switches are connected to core gigabit switch using one or two gigabit uplinks.[citation needed]

Data centers

Google has numerous data centers scattered around the world. At least 12 significant Google data center installations are located in the United States. The largest known centers are located in The Dalles, Oregon; Atlanta, Georgia; Reston, Virginia; Lenoir, North Carolina; and Goose Creek, South Carolina.[12] In Europe, the largest known centers are in Eemshaven and Groningen in the Netherlands and Mons, Belgium.[12] Google's Oceania Data Center is claimed to be located in Sydney, Australia. [13]

Project 02

One of the larger Google data centers is located in the town of The Dalles, Oregon, on the Columbia River, approximately 80 miles from Portland. Codenamed "Project 02", the $600 million[14] complex was built in 2006 and is approximately the size of two football fields, with cooling towers four stories high.[15] The site was chosen to take advantage of inexpensive hydroelectric power, and to tap into the region's large surplus of fiber optic cable, a remnant of the dot-com boom. A blueprint of the site has appeared in print.[16]

Summa papermill

In February 2009, Stora Enso announced that they had sold the Summa paper mill in Hamina, Finland to Google for 40 million Euros.[17][18] Google plans to invest 200 million euros on the site to build a data center.[19] For Google the reason to choose this location was the availability of renewable energy close by.[20]

Modular Container Data Centers

Since 2005,[21] Google has been moving to a containerized modular data center. Google filed a patent application for this technology in 2003.[22]

Software

Most of the software stack that Google uses on their servers was developed in-house.[23] It is believed that C++, Java, and Python are favored over other programming languages.[24] Google has acknowledged that Python has played an important role from the beginning, and that it continues to do so as the system grows and evolves.[25]

The software that runs the Google infrastructure includes:[26]

  • Google Web Server — Custom Linux-based Web server that Google uses for its online services; according to Google, this is not based on Apache.
  • Storage systems:
  • Chubby lock service
  • Borg — job scheduling and monitoring system[28]
  • MapReduce and Sawzall programming language
  • Indexing/search systems:
    • TeraGoogle — Google's large search index (launched in early 2006), designed by Anna Paterson of Cuil fame.[29]
    • Caffeine (Percolator) — continuous indexing system (launched in 2010).[30]

Google has developed several abstractions which it uses for storing most of its data:[31]

  • Protocol buffers — "Google's lingua franca for data"[32], a binary serialization format which is widely used within the company.
  • SSTable (Sorted Strings Table) — a persistent, ordered, immutable map from keys to values, where both keys and values are arbitrary byte strings. It is also used as one the building blocks of BigTable.[33]
  • RecordIO — a sequence of variable sized records.[34][31][35]

Software development practices

Most operations are read-only. When an update is required, queries are redirected to other servers, so as to simplify consistency issues. Queries are divided into sub-queries, where those sub-queries may be sent to different ducts in parallel, thus reducing the latency time.[5]

To lessen the effects of unavoidable hardware failure, software is designed to be fault tolerant. Thus, when a system goes down, data is still available on other servers, which increases reliability.

Search infrastructure

Index

Like most search engines, Google indexes documents by building a data structure known as inverted index. Such an index allows obtaining a list of documents by a query word. The index is very large due to the number of documents stored in the servers.[11]

The index is partitioned by document IDs into many pieces called shards. Each shard is replicated onto multiple servers. Initially, the index was being served from hard disk drives, like it's done in traditional information retrieval (IR) systems. Google dealt with increasing volume of queries by increasing number of replicas of each shard and thus increasing number of servers. Soon they had found that they had enough servers to keep a copy of the whole index in main memory (although with low replication or no replication at all), and in early 2001 Google switched to an in-memory index system. This switch had "radically changed many design parameters" of their search system, and allowed them to enjoy a big increase in throughput and a big decrease in latency of queries.[36]

In June 2010 Google rolled out a next-generation indexing and serving system called "Caffeine" which can continuously crawl and update search index. Previously, Google updated its search index in batches using a series of MapReduce jobs. The index was separated into several layers, some of which were updated faster than the others, and the main layer wouldn't be updated for as long as two weeks. With Caffeine the entire index is updated incrementally on a continuous basis. Later Google revealed a distributed data processing system called "Percolator"[37] which is said to be the basis of Caffeine indexing system.[30][38]

Some details about Google's inverted index compression schemes have been made public.[36][39]

Server types

Google's server infrastructure is divided in several types, each assigned to a different purpose:[11][5][40][41][42]

  • Google web servers coordinate the execution of queries sent by users, then format the result into an HTML page. The execution consists of sending queries to index servers, merging the results, computing their rank, retrieving a summary for each hit (using the document server), asking for suggestions from the spelling servers, and finally getting a list of advertisements from the ad server.
  • Data-gathering servers are permanently dedicated to spidering the Web. Google's web crawler is known as GoogleBot. They update the index and document databases and apply Google's algorithms to assign ranks to pages.
  • Each index server contains a set of index shards. They return a list of document IDs ("docid"), such that documents corresponding to a certain docid contain the query word. These servers need less disk space, but suffer the greatest CPU workload.
  • Document servers store documents. Each document is stored on dozens of document servers. When performing a search, a document server returns a summary for the document based on query words. They can also fetch the complete document when asked. These servers need more disk space.
  • Ad servers manage advertisements offered by services like AdWords and AdSense.
  • Spelling servers make suggestions about the spelling of queries.

References

  1. ^ "Google Stanford Hardware." Stanford University (provided by Internet Archive). Retrieved on July 10, 2006.
  2. ^ Tawfik Jelassi and Albrecht Enders (2004). "Case study 16 — Google". Strategies for E-business. Pearson Education. p. 424. ISBN 0273688405. {{cite book}}: Unknown parameter |isbn13= ignored (help)
  3. ^ [1], april 2009.
  4. ^ Google Surpasses Supercomputer Community, Unnoticed?, May 20, 2008.
  5. ^ a b c d Web Search for a Planet: The Google Cluster Architecture (Luiz André Barroso, Jeffrey Dean, Urs Hölzle)
  6. ^ Strassmann, Paul A. "A Model for the Systems Architecture of the Future." December 5, 2005. Retrieved on March 18, 2008.
  7. ^ Carr, David F. "How Google Works." Baseline Magazine. July 6, 2006. Retrieved on July 10, 2006.
  8. ^ a b Jeff Dean. (2009). Design, Lessons and Advice from Building Large Distributed Systems.
  9. ^ Hennessy, John; Patterson, David (2002). Computer Architecture: A Quantitative Approach (Third ed.). Morgan Kaufmann. ISBN 1558605967..
  10. ^ Randall Stross (2008). Planet Google. New York: Free Press. p. 61. ISBN 1-4165-4691-X.
  11. ^ a b c Fiach Reid (2004). "Case Study: The Google search engine". Network Programming in .NET. Digital Press. pp. 251–253. ISBN 1555583156. {{cite book}}: Unknown parameter |isbn13= ignored (help)
  12. ^ a b Rich Miller (March 27, 2008). "Google Data Center FAQ". Data Center Knowledge. Retrieved 2009-03-15.
  13. ^ Brett Winterford (March 5, 2010). "Found: Google Australia's secret data network". ITNews. Retrieved 2010-03-20.
  14. ^ Google "The Dalles, Oregon Data Center" Retrieved on January 3, 2011.
  15. ^ Markoff, John; Hansell, Saul. "Hiding in Plain Sight, Google Seeks More Power." New York Times. June 14, 2006. Retrieved on October 15, 2008.
  16. ^ Strand, Ginger. "Google Data Center" Harper's Magazine. March 2008. Retrieved on October 15, 2008.
  17. ^ "Stora Enso divests Summa Mill premises in Finland for EUR 40 million". Stora Enso. 2009-02-12. Retrieved 12.02.2009. {{cite web}}: Check date values in: |accessdate= (help)
  18. ^ "Stooora yllätys: Google ostaa Summan tehtaan". Kauppalehti. Helsinki. 2009-02-12. Retrieved 2009-02-12.
  19. ^ "Google investoi 200 miljoonaa euroa Haminaan". Taloussanomat. Helsinki. 2009-02-04. Retrieved 2009-03-15.
  20. ^ Finland - First Choice for Siting Your Cloud Computing Data Center. Accessed 4 August 2010.
  21. ^ http://www.theregister.co.uk/2009/04/10/google_data_center_video
  22. ^ http://patft.uspto.gov/netacgi/nph-Parser?Sect2=PTO1&Sect2=HITOFF&p=1&u=/netahtml/PTO/search-bool.html&r=1&f=G&l=50&d=PALL&RefSrch=yes&Query=PN/7278273
  23. ^ Mark Levene (2005). An Introduction to Search Engines and Web Navigation. Pearson Education. p. 73. ISBN 0321306775. {{cite book}}: Unknown parameter |isbn13= ignored (help)
  24. ^ http://www.artima.com/weblogs/viewpost.jsp?thread=143947
  25. ^ http://python.org/about/quotes/
  26. ^ http://highscalability.com/google-architecture
  27. ^ a b c Andrew Fikes. Storage Architecture and Challenges. Google TechTalk. July 29, 2010.
  28. ^ Intel Corp. Seizing the Open Source Cloud Stack Opportunity. See slide "Proprietary Cloud Computing Stacks".
  29. ^ Anna Patterson - CrunchBase Profile
  30. ^ a b The Register. Google Caffeine jolts worldwide search machine
  31. ^ a b http://www.eweekeurope.co.uk/news/news-it-infrastructure/google-developing-caffeine-storage-system-1620
  32. ^ http://code.google.com/apis/protocolbuffers/docs/overview.html
  33. ^ http://labs.google.com/papers/bigtable-osdi06.pdf
  34. ^ http://www.windley.com/archives/2008/06/velocity_08_storage_at_scale.shtml
  35. ^ http://groups.google.com/group/protobuf/browse_thread/thread/ee27572aef9da70a
  36. ^ a b Jeff Dean's keynote at WSDM 2009
  37. ^ Daniel Peng, Frank Dabek. (2010). Large-scale Incremental Processing Using Distributed Transactions and Notifications. Proceedings of the 9th USENIX Symposium on Operating Systems Design and Implementation.
  38. ^ The Register. Google Percolator – global search jolt sans MapReduce comedown
  39. ^ GroupVarInt encoding from Jeff's talk is also described in the paper: Boulos Harb, Ciprian Chelba, Jeffrey Dean, Sanjay Ghemawat. (2009). Back-Off Language Model Compression. Proceedings of Interspeech 2009, pp. 325-355.
  40. ^ Chandler Evans (2008). "Google Platform". Future of Google Earth. Madison Publishing Company. p. 299. ISBN 1419689037. {{cite book}}: Unknown parameter |isbn13= ignored (help)
  41. ^ Chris Sherman (2005). "How Google Works". Google Power. McGraw-Hill Professional. pp. 10–11. ISBN 0072257873. {{cite book}}: Unknown parameter |isbn13= ignored (help)
  42. ^ Michael Miller (2007). "How Google Works". Googlepedia. Pearson Technology Group. pp. 17–18. ISBN 078973639X. {{cite book}}: Unknown parameter |isbn13= ignored (help)

Further reading

External links

Retrieved from "https://en.wikipedia.org/w/index.php?title=Google_data_centers&oldid=423467856"