Google Mini Search Appliance review: Google Mini Search Appliance

Google Mini

The Google Mini lets you set up a custom search engine in minutes, allowing you to index and search up to 100,000 documents in more than 220 different file formats, including HTML, PDF, and Microsoft Office file types. We definitely recommend the Google Mini if you need an easy-to-use way of indexing and searching a relatively small collection of Web pages. It's almost trivial to set it up and get started, much easier than having to set up your own separate hardware and search software. The hardware is simple enough: a bright blue 1U-height rack-mount box with a big Google logo painted on top. Connect a network cable (included) and the power cord, then wait for the happy tone that tells you it has finished booting up. There's no separate power switch--once you've plugged it into a wall outlet, it begins the boot-up process. At this point you have to do the network configuration. That means giving the Mini its IP address for your network, telling it what e-mail server to use when it sends out status alerts and such. To do this, you must plug a laptop into the Mini or haul the unit over to a desktop.

There are two RJ-45 ports in the back of the Mini--a network port that connects it to your regular network and a configuration port designed specifically for the initial configuration tasks. Google provides a special crossover cable that lets you plug your laptop into that configuration port, and the company has thoughtfully put a tag on that cable telling you what the config IP address is. You simply fire up a browser on the laptop and point it at the Mini's config IP address. At this point, your Mini needs to be told how to use the network port--what IP address it should be, what DNS machines to query, what SMTP server to use, and so on. The Mini's administration console is essentially a series of Web forms where you enter this information.

The Google Mini can obtain some of the needed configuration info from your network's DHCP server. However, you will need to give the Mini a static IP address (provided by your network admin) and a SMTP server address, which isn't delivered via DHCP.

Test functions built into the admin console help you confirm whether you've input valid addresses and names. We may have been a bit lucky, but the time from the moment we first plugged in our Mini to the time the initial configuration was finished was about 30 minutes. Our one complaint is that the configuration port is in the back. In theory, you'll probably never need to change the network configuration, but we would like to see a configuration port in the front just in case. Depending on your rack space, it could be a little complicated to get a laptop reconnected to that rear configuration port.

Once you've configured the network, you can unplug the laptop and perform the actual search configuration from any browser on your network. The simplest configuration is to just specify a name for your collection, give it an initial URL where it should begin the crawl, then accept all of the other default parameters. If you go that route, you could have it beginning the crawl within a few minutes.

However, you'll probably want to use at least a few of the other search options. The Mini can index a maximum of 100,000 documents, so the most likely parameters you'll want to control are the URL patterns you want it to crawl. That way you don't waste any of those 100,000 documents crawling URLs that you don't need. For example, if your site provides printer-friendly versions of all Web pages, you'll probably want the Mini to ignore them. You do all this through the Google Mini's browser-based console. The UI is a bit bare, but it's also self-explanatory. One helpful item for those of us who don't live and breathe regular expressions: there's a test button that lets you see whether a specific URL will fit the patterns you've input.

Once the Mini has finished the crawl, it builds an index from all the documents, then copies that index out to be served from a query interface. The query interface looks, of course, an awful lot like Google.com. If the Mini is running as part of your intranet, you can stick with the basic customizations, such as using your company logo instead of Google's at the top of the page. If you'll be using the Mini to serve up a public search page, you'll want to use the built-in XSLT editor to really customize the pages so that they look like the rest of your site.

We also like the Synonyms feature. If, for example, you have a lot of Macintosh content, you can specify the word Macintosh as a synonym for Apple. When users search for Apple, they'll see a hint at the top of the search result page telling them they can also search for Macintosh. Click Macintosh, and the Mini will carry out the search.

One thing we find a bit annoying is that Synonyms are strictly one-way. If you need a pair of words to be synonyms in both directions, you have to explicitly set up both. Google says this is for maximum flexibility, and that's probably true, but it's still a little irritating for those of us who expect synonyms to work automatically in both directions.

Another search option we like is KeyMatch; for a specific search word or phrase, you can specify a URL and text to display at the top of the results page. It ends up looking and working a bit like the keyword ads on Google.com. If you're having a special sale on soap, for example, you might set up a KeyMatch so that whenever users search for soap, they see a short note at the top of the page and a link that takes them to the sale page.

Unfortunately, the upgrade path poses problems. If the Mini's 100,000-document limit ever becomes an issue, you could buy a second Mini. But that means having to find a way to configure the two machines so that they crawl separate sections of your site, and it means users would have to query both interfaces to search your site. Your other option is going for the big Google Search Appliance, which can index anything from 500,000 documents on up. But it starts at a $30,000 price point, a substantial jump from the Mini's more reasonable $2,995 price.