Attivio | Unified Information Access Blog

I was delighted to attend Gilbane 2008 at the Westin Copley in Boston recently. There were many interesting sessions, and a few spirited exchanges with the audience, particularly when one of the keynotes included offensive pictures in his presentation, and another when an analyst from Forrester bravely predicted that Twitter is just a flash in the pan.

One of the conference highlights was the "Plug-and Play: Enterprise Experiences with Search Appliances" session. Angela Foster, IT Manager at FedEx.com and Dennis Shirokov, Marketing Manager, FedEx Digital Access Marketing, FedEx, spoke about their experiences implementing an appliance to power the search for Fedex.com. In summary:

• On a slide titled: "Be Ye Not So Stupid", Ms. Foster listed warnings that included:

  1. The Account Executive just wants to make the sale
  2. The Box is not magic
  3. There's no such thing as a free lunch
  4. Cool does not equal competent
  5. Do not unplug before shutdown
  6. Got Information Security?
  7. There is no such thing as Plug and Play
  8. If they say Plug & Play then recommends an integration specialist, see number 7
  9. What You See Isn't Always What You Get (WYSIAWYG)
  10. Wait, I thought this came with.....

• She said that their choice of an appliance was a "mistake", and that the implementation was painful overall.

• As a team they were overly "dazzled" by the sales and marketing people from the appliance provider, which has a very popular brand, and didn't look skeptically enough at the claims.

• They put too much importance on (low) price and not enough effort into understanding the requirements.

• Configuration management was apparently a huge problem, with all configuration changes having to be made manually, for each server.

• Managing relevancy was a huge challenge; in particular manual adjustment of relevancy for various keywords was simply not supported.

• Some of the simple things they were told were possible required significant coding; when pressed on this issue, the vendor recommended consultants to help script the desired behavior.

• Many of the things they were told were possible were simply not practical.

• Amazingly, one of the few positive things FedEx reported is that the appliance is only supported by the manufacturer for three years, so they'll be forced to buy new hardware and not hang onto legacy hardware.

I wonder if that last point was factored into their TCO. At any rate, it's not exactly a ringing endorsement, is it? I've been attending conferences and listening to customer implementation stories for years and I've never heard such a negative take on a vendor.

My two cents: FedEx is right, the appliance is the wrong choice for building a public-facing information portal. It was designed to focus on a fairly trivial use case - crawling and indexing intranet content and providing end users with a single result list that includes external web results. It is, by design, incredibly easy to use, and extremely low priced. The trade off is that it is, by design, neither sophisticated nor customizable. In the intranet world, who cares if result #3 is the best one? That's good enough. But that standard clearly won't fly on a world class site like Fedex.com.

I was also happy to speak on the Open Source Search Applied in the Enterprise panel with Leslie Owens, Analyst, Forrester Research, and Lynda Moulton, Lead Analyst for Enterprise Search, Gilbane Group. Some highlights:

• Forrester has extensive survey results that shows primary adoption of open source search among ISVs and SaaS/ASP players - she reported little (and slow) adoption of free open source software (FOSS) search in other enterprises.

• The survey further shows the big issues for adoption are related to risk: better history buying vs. building, concern about implementation cost/complexity, and licensing issues.

During my presentation I argued that the origin of a technology often helps predict how it will evolve. For example, one of the leading enterprise search vendors started out as a web search vendor. Not surprisingly their technology was successful in the public-facing portal market. The FOSS search "engines" were created by developers for developers, and as a result they are much loved by developers, but hard for others to use. (This is how open source got started; it was all about technology: operating systems like Linux, Apache webserver, etc.)

Another key point is that building a really good search application is hard ... not everyone has the expertise to do it, and many different technologies can be integral to success. If you are comfortable building/integrating software, and especially if you want to build something unique - as many ISVs are - looking at FOSS makes sense. If not, you may want to consider commercial products, or perhaps a combination of FOSS, commercial software and services.

Finally, I noted that the main FOSS search communities are very vital and enthusiastic, spinning out projects to address some of the technology areas where FOSS falls short of more complete commercial offerings. So the future is bright. The relational database took decades to become an inevitable part of enterprise infrastructure; search is considerably less mature and solved, but between commercial and FOSS efforts and customers with real problems to address, evolution will continue.

Thanks to everyone at Gilbane for putting on such a great show. I look forward to attending next year.

TrackBack URI for this entry

Comments (0)