Google or bust?

September 16, 2005 8:07 AM

Does anyone have any experience with the Google Search Appliance or similar enterprise search tools?

In particular, I'd like to know how much I can customize the external search functionality. For example, my users will have the option of searching my site and/or the web, but if they search the web, I'd like to restrict the sites that the search returns/filter for keywords, that sort of thing. Is this possible? I also want to customize the interface to the point where an end user won't recognize it as a Google product. The FAQ seems quite vague, and Google doesn't want to talk to me. Besides that, any general impressions or alternate recommendations? Thanks...

posted by loquax to computers & internet (9 comments total)

Anandtech recently switched to Google Search, their description might be interesting to you.
posted by bering at 8:55 AM on September 16, 2005

I've got plenty of experience, as I'm our system administrator for two google boxes (one regular, one mini).

The Google box does not let you route regular Google searches through it - you cannot search "the web" through it. You have to explicitly spider the sites you want in your index. Granted, those sites don't have to be on your network, per se, but other administrators might get pissed if you pound their server into oblivion.

In terms of customizing the interface, yes, this is definitely doable through XSLT (and there's also some more basic hand-holding tools you can use). You can check out ours and see that it looks nothing like a Google search.

Overall, I've been very happy with both the GSA and the Mini in terms of administration and ease of use. There's routinely updates for it to fix bugs and add functionality, they recently increased the document limits on everyone's licenses, etc etc. Documentation is good, haven't had to go to their support yet (worth noting, although I think this is in the FAQ: there is no phone support, only email).

Any other questions?
posted by Remy at 9:29 AM on September 16, 2005

Ditto GarageGames.
posted by phearlez at 9:30 AM on September 16, 2005

At my last job we looked at Google Appliance and Thunderstone and went with the latter because it offered more control "under the hood."

Google offers "webinars" where many many potential customers can see Google's presentation. Thunderstone met with us (via conference call) one-on-one, which Google wouldn't do.

YMMV
posted by terrapin at 10:17 AM on September 16, 2005

Thanks for the answers so far all, I'll check out Thunderstone as well, but I think I'd need something closer to the full Google appliance rather than the mini, not sure if Thunderstone offers a bigger appliance.

Remy, are you saying that the full Google search box (not the mini) does not allow you to have a general web search in addition to a site search? I'd like to provide two sets of results to our users, one from our site, and one from "the web". Is this not possible? I assume that the standard free web search (like the metafilter one) comes with Google ads, which I would want to avoid.

Also, is it possible to tweak or modify the search algorithms used by Google search (I'm guessing not)? Or is the customization of the search limited to keyword association and site filters? If it is locked, have you found that the native search works well with your content? We tried other third party search software providers and have had major problems with the page rankings and relevancy.

Thanks again...
posted by loquax at 11:52 AM on September 16, 2005

loquax: It's not too hard to write something up in PHP or Perl that will do what you're looking for. It's a matter of passing the search string to Google using the Google API, and then formatting the results that are given back to you.

It's not something you need the search appliance for (for searching the web, that is)
posted by SweetJesus at 1:14 PM on September 16, 2005

Err, that's for grabbing web results and dispaying them along with your internal results. I don't think you can do anything about the Page Rank algorithm, and the way it works (but I could be wrong).
posted by SweetJesus at 1:16 PM on September 16, 2005

I had the opportunity to work with a Google Search Appliance a couple of years ago. Unless they have changed things you can not do anything with the search algorithms. You can modify the interface like others have described above.

Their purchasing and licensing scheme may have changed lately but when my organization bought one they paid I think around 35k for it. This was at a college so we may have gotten special pricing. The whole Google appliance was only good for like a year it might have been two I can't remember. I mean everything about it was only good for a year. Hardware ,software and everything. You had to configure the appliance to synch up with a time server and after a certain date the appliance simply quit working. You could not extend the time on your license. If you wanted to continue using a Google Search appliance you had to buy a whole new one hardware and all. You might have been able to configure it to synch with your own time server that you kept backdated. But of course that would be illegal.

We did not elect to purchase a new one when ours died. We turned it over to our Unix guys to see if they could find a way to hack into it and make something usefull out of it. Not sure if they ever did.

Google not in spellchecker?
posted by Justin Case at 2:39 PM on September 16, 2005

- No, there is no external web searching. What you spider is what you can search.

- You cannot tweak the algorithm, outside of keyword matches and things like that.

- The native search works very well.

- Justin's information about the license expiring and the box doing so as well used to be true, but is no longer - now when your license expires, it continues to run, you just won't be able to get support on it.
posted by Remy at 5:57 AM on September 18, 2005