Kal Kan - You have NO CLUE do you ?
Every hit on GOOGLE is generated with THOUSANDS OF Intel-based SERVERS !!!
"To deliver whip-fast search results from such a massive database, Google employs a vast server farm of more than 3,500 uni-processor servers based on Intel Architecture processors and linked using Intel® PRO/100 Ethernet adapters. "
What a HOOT you are !!!
intel.com;
Google, Inc., is one of the most innovative companies on the Web. Its unique Internet search engine technology and infrastructure makes fast work of even the most laborious Web searches. Founded in 1998 by Stanford University PhD candidates Larry Page and Sergey Brin, Google has quickly and quietly emerged as one of the most popular search engines on the Web.
Despite having a limited advertising and marketing budget, Google (www.google.com and its co-branded partners) fields more than 13 million searches every day—up from just 500,000 per day in June 1999. Currently the search engine of choice for Netscape's* popular NetCenter* Web site, the company also offers free and fee-based Google-powered Web searches to third-party Web sites.
Searches Made Fast
If Google's business could be summed up in one word, it would be focus. While Web search behemoths like Yahoo!* and AltaVista* build portals and roll out e-mail, chat and other services, Google continues to refine and extend its core search technology. The single-minded focus is starting to pay dividends in the form of lucrative OEM deals, escalating usage, and excellent word-of-mouth buzz.
A privately held company, Google earns money by licensing its technology to customers like Netscape NetCenter, Red Hat*, Washington Post and Virgin Net* as well as by displaying advertising on search results pages. True to the company's mission of delivering fast searches, its ads consist of simple text.
One thing is certain, if you want your search results fast, you can do a lot worse than Google's high-speed service. Benchmark results by independent publications consistently show Google to provide the fastest searches on the Web. What's more, Google's unique way of assessing the value of results—by the number of links pointing to a site—seems to produce eerily relevant returns. The result: Users who come to Google.com once tend to come back again and again.
In fact, Google was named to Time magazine's "Best Cybertech of 1999" list and ranked first in Time Digital magazine's "Top Ten Sites" in 2000. Google also received a 1999 PC Magazine Technical Excellence Award. The search engine also twice ranked first overall among 13 leading search and portal sites in surveys of user satisfaction and loyalty conducted by NPD Online Research in the last quarter of 1999 and first quarter of 2000. Nevertheless, as Google has won over customers, it has had to contend with breakneck growth.
Of course, all the accolades won't mean much if Google bankrupts itself on expensive proprietary hardware. Jim Reese, chief operations engineer at Google, says that Intel® Architecture servers and the Linux* operating system fit Google's business plan. "With Linux running on the Intel® platform, there is no better price/performance that we could find. When all is said and done, that is the bottom line."
Off the Shelf, Vastly
Google may be the ultimate scale-out proposition. The company employs redundant arrays of inexpensive servers (RAIS) to tap into an index between one and two terabytes (1,000 gigabytes) in size and consisting of over 200 million Web pages. To deliver whip-fast search results from such a massive database, Google employs a vast server farm of more than 3,500 uni-processor servers based on Intel Architecture processors and linked using Intel® PRO/100 Ethernet adapters. The servers are outfitted with 256MB to 1GB of RAM and run the Linux operating system and a suite of custom-built applications.
Google's application makes expensive proprietary hardware unsuitable, says Reese. "We are not like a transaction-based e-Commerce site, where it makes sense to spend a whole lot of money on some really big server iron and storage area network. We architected our solution to be scalable by using smaller servers that are multiply redundant and very fast through load balancing. Also it makes us very fault tolerant—we can lose a whole cluster or clusters, and we'll still be fine."
In Google's environment, disk I/O performance is an overriding factor, yet the cost of high-end SCSI disk subsystems is prohibitive. So the company standardized around inexpensive IDE technology, outfitting each server with a pair of internal disks storing either 22GB or 40GB apiece. "We did a lot of benchmarking early on, and we found that for the best price/performance we would set up two IDE hard disks, each on a separate controller," says Reese.
Google's massive search index is distributed and mirrored among the approximately 7,000 individual disk drives, enabling custom load-balancing software to point queries at the most-available servers and disk subsystems. Network connectivity is enabled through Intel PRO/100 Ethernet adapters at the servers linked to a gigabit Ethernet backbone. The result: Peerless search response.
By employing redundant arrays of inexpensive servers (RAIS), Google enables a vastly scalable, economical, and robust infrastructure that can adapt to surging demand. The company estimates that it adds approximately 30 new servers a day to its farm, just to keep pace with mounting demand. The servers are split among a pair of co-location facilities in the San Francisco Bay Area and a third facility on the East Coast. The company is also considering facilities in Asia and Europe as a way to cut down the latency for searches performed by users in those regions.
Tailoring Solutions
With 10,000 servers expected to be installed by the end of the year, co-location costs are an overriding factor. From the beginning, Google worked with several OEMs to obtain uniquely compact server form factors and cabinets. The company says it can fit 80 servers in a space 7 feet tall by 2 feet wide and 2.5 feet deep. Hardware is sourced from numerous vendors to ensure timely delivery. Rackable Systems*, an OEM specializing in compact servers, builds the systems.
Of course, managing such a large farm of servers presents a unique challenge. The company crafted its own solutions for operations such as remote management and load balancing. Google built much of its own software, including remote management consoles and load-balancing engines. It has also developed a streamlined method for configuring servers that enables new systems to be brought online very quickly.
"We have basically made our machines modular. We start with the basic hardware and the base Linux OS, and then we 'Google-ize' it, if you will, with the base Google code that any machine might need," explains Reese. "At that point, any machine can be made into any other machine. So if we need to add Web servers, any machine can be made into a Web server, because all the machines have identical configurations."
Google continues to refine its services as well. The GoogleScout feature added to search results enables visitors to deepen their searches by Web pages related to a specific search return. The "I'm Feeling Lucky" button on each search page takes the user directly to the page of the top-ranked result in a search. Google's highly modular and rapid system configuration process enables the company to scale its infrastructure as new services like these are deployed.
With over 3,500 Intel Architecture servers already humming away at Google's co-location facilities, the question is how many more servers will the company end up installing? At this point, the sky seems to be the limit. Google says it has been growing at a rate of 25% per month, and as more and more Web sites employ Google's free and fee-based Web search service, that rate could accelerate.
One thing is certain, the company is ready, willing and able to bring additional Intel Architecture servers to bear on the challenge. Google has crafted a vastly scalable infrastructure that makes the most of its unique and sophisticated search engine technology. As traffic at www.google.com continues to grow by leaps and bounds, the company will simply add more servers to maintain Google's edge as one of the fastest search engines on the Web.
Related Information |
