Google upgrades Search Appliance

Google Inc. has upgraded its Search Appliance, improving the capacity and performance of the device, which combines hardware and software to provide in a box the search functionality employed by the Web site.

Introduced in early 2002, the appliance is aimed at companies, educational institutions and government agencies that want to make their sites searchable using Google technology. Thus, a university might buy a Google Search Appliance to provide search capabilities for its student and employee intranets, as well as its public Web sites.

The appliance has been revamped to index more documents, do so more intelligently and perform more queries per minute, said Dave Girouard, Google's enterprise unit general manager. The new version of the product also features improved security and allows for collections of indexed documents to be partitioned with more flexibility and granularity, he said.

"We launched this product quietly in 2002 and it has grown nicely and become a successful business for Google," he said. "This is our first major new upgrade of the product."

In terms of performance enhancements, the new version can index as many as 1.5 million documents, which is five times as many as the first version, and execute 300 queries per minute, also a fivefold improvement, Girouard said.

The new version also features more intelligent and efficient document crawling. The first version crawled documents in batch fashion, meaning it would scan and index the entire collection of documents every time the administrator scheduled a refresh. The new version scans and indexes only those documents that have changed since the last crawl, an improvement that speeds up the process and reduces consumption of bandwidth and processing power, Google said.

In addition, administrators don't have to schedule the updates, since the new version is continuously crawling the collection, which results in changes being indexed more promptly. Thus, with the first version, the Search Appliance would be configured to run a batch update once a day, or once every two days, which could delay changes until the update was run, while the new version detects changes soon after they're made, Girouard said.

Google also enhanced the product's security by improving its ability to prevent users from viewing documents they're not authorized to access, he said. After executing a query, the upgraded product rounds up all the documents that contain the keywords and then filters those documents based on the user who made the query, showing only the documents that the user has permission to view, he said.

Another new feature is the ability to create different collections of documents, whereas the first version allowed for the creation of only one collection of documents, he said. Thus, with the new version, a company might create a collection of searchable documents for its sales and marketing employees, a different one for its call center employees and so on.

A related new feature is the product's ability to support different user interfaces for a single collection. Thus, the administrator might set up a user interface for the sales and marketing employees that is different from the user interface for the call center employees, while having both sets of users access the same collection of documents, he said.

The new version of the Search Appliance is twice as tall as the first version because it has more powerful hardware, which in turn generates more heat and requires more space for cooling, Girouard said. That means it is 3.5 in. (2U) high and 19 in. wide.

Google doesn't reveal which vendor makes the appliance's hardware. "It's commodity hardware -- the same general hardware we use in our Google data centers," Girouard said.

A basic installation of the Search Appliance can be completed in as little as 30 minutes, allowing an IT department to have it up and running in a matter of hours, he said. Installations that involve deeper customization will take longer to complete.

The Search Appliance can crawl and index documents in more than 250 file formats, as long as the documents are accessible via HTTP, Girouard said. It supports 28 languages, and search formats such as natural language, keywords and Boolean, he said.

It delivers cached page results, allows for document sorting by date, lets users search within results, and features a self-learning spell checker that suggests alternate spellings for queries. For administrators, the product generates usage reports and crawl analysis.

The product is sold as a stand-alone device under its GB 1001 model number. A GB 1001 with a capacity of 150,000 documents starts at $32,000, while one with the maximum capacity of 1.5 million documents costs $175,000, Girouard said. The new version of the GB 1001 is available now. Included in the price are two years of customer support, he said.

The Search Appliance is also sold in preconfigured stacks of multiple GB 1001s. The GB 5005 is a stack of five devices, while the GB 8008 is a stack of 12 devices. (In the first version, the GB 8008 was a stack of eight devices.) Google preconfigures these stacked devices to work together, Girouard said.

At press time, Google didn't have statistics for the performance improvements that the stacked products of the new version will offer, Girouard noted. Prices for the stacked products are determined by the number of documents they're able to index, he said.

Copyright © 2004 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon