Exalead

An Interview with François Bourdoncle

Exalead's office is located in Paris, a short walk from the Champs-Elysées and the Rue de Rivoli. Finding my way to 10 Place de la Madeleine was an excellent prelude to my interview with François Bourdoncle, the youthful founder of Exalead. I could tell that Exalead's business was in high gear. Staff mobiles were chirping, and two well-fed, 55-year-old "serious" executives in suits waited for a product demonstration. I learned about Exalead three years ago when I was looking into AltaVista.com, the one-time Web search leader. Mr. Bourdoncle's name in the roster of wizards who labored on that system in the late 1990s, well before the Compaq - Hewlett-Packard egregious mishandling of AltaVista.com

I have been watching Exalead expand its product offering from a Web search system to a comprehensive, extensible enterprise solution, then to product innovations in the eDiscovery sector and social systems. Mr. Bourdoncle is a computer scientist's computer scientist. Like Jeff Dean, Sanjay Ghemawat, and the other AltaVista superstars, Mr. Bourdoncle could be working in Google's technical vineyard Messrs. Brin and Page.

Mais non!

Instead, Mr. Bourdoncle has labored for a decade to create what may be the search system ideally positioned to challenge Google and the leaders in the behind-the-firewall / enterprise search battleground. Exalead now has clients in more than a dozen countries, operates its own data centers so licensees can exercise a hosted solution, and makes the phrase "Search by Serendipity" a buzzword in the search community. I interviewed Mr. Bourdoncle in February 2008.

Paris has become a center of innovation in search, information retrieval, and content processing. There's your firm, Exalead. Polyspot, Kartoo, Lingway, and others. Is it French wine or something else?

Wine may have something to do with it, you’re right. But wine produces different effects on different people -- makes some giddy, brings others down. So you should not rely on it too much to design your products!

More seriously, it is true that France has a great tradition of designing programming languages, which I did in my past life as a researcher. For instance, ADA (the language mandated by the US Department of Defence) was designed by a French team.

France also has a strong tradition of working on linguistics and semantics. It is absolutely logical that search engines are an area of interest to French technologists. However, Exalead’s focus on core infrastructure is fairly uncommon in France, if not unique. Our Exalead engineering has more to do with my past life as a researcher at Digital Equipment Corp. in the US than my being 100 percent French.

What was the trigger in your career that made search and retrieval a focal point? Weren't there other, easier opportunities for you to use your technical training and expertise?

As a general rule, I do not like doing easy things, since they tend to become fairly boring very quickly. I am a creative person, and I like to explore new territories. Like Frank Sinatra said, "I want to do it my way." 

This is probably much more American than French, I have to admit, but I certainly have some cultural flexibility in that respect. The search trigger for me was a phone call. Louis Monier -- AltaVisa’s founder and, as you know, also French -- said, "Hey, how are you doing? We’re starting a new project. It’s called a search engine to help you find stuff on the Web. You have to come back to the US and help us. It’s going to be a lot of fun”.

That was in the fall of 1995. I took the first plane I could catch to the US. Search offered a way to capitalize on my interests, rather than simply doing one thing and then another -- always solving new problems, always learning.

In fact, amusingly enough, I’ve even managed to use the programming language that I had designed after many years of research.

This is our ExaScript. It uses a Java-like syntax, the XML data model, an inline syntax for XML objects similar to PHP, and a multiple dispatch mechanism similar to Python’s. ExaScripts does all this with the most sophisticated higher-order polymorphic-type system in use for any practical language today.

So you share some DNA with AltaVista.com and, by extension, with Google. What did you learn from your work with AltaVista.com?

Very fundamental things. For example, how to write server-side software (the 24x7 software that never breaks), how to write memory-efficient programs, how to use 64-bit processor architectures (flat address space, extreme processing power), how to design software compatible with what is now called "the grid-computing model". But we didn't have a good name like that for the technique then, however.

I also learned how to write reliable and efficient multi-threaded programs on multi-processor systems. You have to remember that the first 64-bit processor was Digital Equipment Corp.’s Alpha chip, and I was lucky enough to have had one of the first prototype workstations to play with. It had an unexpected benefit.

What was that?

It would also heat my office. The Alpha was like a tiny furnace.

What other impacts did AltaVista.com have on Exalead's engineering?

At Exalead, from day one, we focused on multi-threaded, 64-bit architectures. But there were not many 64 bit chips in use in 2002. So, our engineering advances were before their time. In those early days at Exalead, most of our customers used the old 32-bit hardware and operating systems. We had a stallion, and we had to make him behave like a child's pony.

Then, the 64 bit chips became available and started to become the standard in many organizations. Exalead's performance was immediately evident. I know that our competitors discovered that it was their turn to suffer.

Today, Exalead has, I know, the most mature, robust and scalable search software. We make full use of today’s multi-core processors. Our products are also able to adapt automatically to various memory / processor / disk configurations. I think you know exalead one: desktop search. It runs on small laptops and on high-end multi-processor servers. But you agree that it's very fast on any platform, right?

Yes, your Exalead system struck me as extremely fast, lightweight, and scalable. But your focus on core architecture engineering is unusual, isn't it?

Architecture is very important, and customers and competitors are now learning this fact the hard way.

The reason why I chose nine years ago to put so much effort on core architecture engineering is that I was convinced that enterprise and Web search were ultimately going to become one market, and that a unified, infrastructure-level software platform was needed to address those two markets.

As it turns out, I was right, as proven by Microsoft’s !.2 billion bid for FAST Search & Transfer, and there is also Google’s push into the enterprise market. The interesting thing for us is that Exalead is now the sole independent vendor to address both enterprise and Web search markets. This is going to be a major asset in the very near future when Business Intelligence meets Competitive Intelligence. This fusion is going to be the next big thing in information processing, I think.

So now Exalead is definitely the fastest and most lightweight and scalable system on the market today. We’ve reduced the average footprint by a factor of 10 versus our closest competitor. This core engineering effort makes a huge difference on high-demand search applications when our clients start deploying hundreds of servers.

As I said, the same binary runs on servers and sub notebooks without any major performance problem, so it is lightweight. And as far as scalability goes, we operate a large-scale Web search engine (www.exalead.com) than indexes eight billion Web pages, making it the Exalead largest Web index after Google, Microsoft, and Yahoo.

Exalead provides point and click options for users. Users seem to want search systems to be easier to use. What are you doing to make search easier for users?

One of the reasons Exalead invests in Web search, beyond the BI / CI convergence I’ve just mentioned, is that it gives us first-hand experience designing innovative user interfaces for accessing information.

The linear search experience that Google has imposed as a de facto standard is hitting a hard limit in the enterprise. In organizations, data are much more structured. Using meta-data to navigate in search results is a must. That is why we have patented a method called “search by serendipity." I know you have heard me use this phrase before.

When we say "search by serendipity," we are talking about our  particular form of assisted navigation. Some people also call this guided navigation or faceted search. The idea is that users can refine their searches and move around the information using both automatically extracted concepts and existing metadata or pre-built categories. This double-sided method is deployed in all of our products and customer implementations. Its universality is now very clear and it can be used in B2C [business to consumer] and B2B [business to business] implementations equally well.

Can you give me an example of a Exalead customer who has used the Exalead API to integrate your system into a third party application?

Our OEM strategy allows our customers to use our search engine in third party applications.  The exalead one:enterprise 4.6 OEM Edition is an embeddable search solution specifically designed for independent software vendors in data- and transaction-intensive markets who have to meet stringent performance, availability and scalability requirements. Seamless integration is pretty standard for us in enterprise content management, business and competitive intelligence, document and records management, e-mail and compliance, enterprise resource planning, and publishing, among others.

For example, Tera Digital Publishing, a content management and archiving solution provider for newspapers, has decided to migrate to exalead one:enterprise 4.6 OEM Edition.

In addition to our OEM clients, some of our other customers integrate Exalead's search engine in internally developed business applications. For example, Gefco SA, a logistics company, is currently developing a new track- and-trace applications based on Exalead's capabilities.

I think the Exalead interface has been changed in the last two or three years? Compared to some other aspects of the Exalead system, how much time and effort do you put into the user interface?

More and more effort is put into the user interface. To give you an example, if you look at our latest B2C product called BAAGZ, which goes into public beta this week [February 25, 2008], you will realize that we are pushing the envelope in terms of UI.

We listen to our users, and we believe that the time for simplistic and “naked” search engines is over, fini. Now is the time for full-fledged “search products,”not simplistic “search engines”. Think of the difference between a car’s engine and the car itself. BAAGZ, which some of the alpha testers have seen, was described as "a new form of search-inspired social networking". Another alpha tester called it  "a new form of social networking-inspired search". BAAGZ is, for the record, the first social network to allow people to connect because they have shared interests.

Inside Exalead, we like to think of it as the first live network of shared interests. And of course, there’s Exalead’s Web semantic search technology at its very core, as you would expect from us.

In 2006, Autonomy bought Verity, worrying the Verity customers about the future of the K2 system. Now Microsoft has acquired Fast Search & Transfer. What's the impact on the search industry of these two big deals?

There’s a logic to these high-profile buy outs. Search is becoming a new form of infrastructure software, just like databases and application servers. Search systems are now increasingly being used as “data servers” if you will.

So it’s logical that large infrastructure vendors want to add search technology to their portfolios. Most of them have tried to do it internally, but these companies have discovered the hard way that building that kind of infrastructure software is really hard, and so they are starting to buy pure players.

And I think this consolidation is pretty good news for us, since we shall soon be the sole independent provider of this kind of infrastructure software. There is a huge market for an independent vendor like Exalead. Combined with our technology and system architecture, these deals are making life interesting.

When you think about behind-the-firewall search (I don't like the term enterprise search), what are the major challenges that Exalead and others have to address?

I think that ease of deployment is key here. You really want to have the best of both worlds: complete flexibility in the kind of data you index, how you can restructure it, etc. and a plug-and-play, no-hassle configuration system. This is not easy, but this has been Exalead’s focus for a long time. We are also working on providing a set of interoperable vertical modules that our enterprise customers can snap together to assemble their specific type of vertical search solution in a matter of days, not months.

Is Exalead going to become a publicly traded company, sell out, or continue to grow organically.

We are definitely growing organically and the future is looking very bright. Did you know that we multiplied our workforce by two last year, our revenues by three, and we signed four times more contracts than the year before, a little more than $15 million.

I am more optimistic than ever that there is a future for Exalead as a very profitable independent company. Time will tell if we want to go public or not.

ArnoldIT Comment

In Beyond Search: What to Do When Your Search System Doesn't Work, I count Exalead as one among four vendors that are candidates to replace an incumbent system. Furthermore, I identified Exalead as an "up and comer," a designation that means a company has strong growth and exceptional technology. My analyses show that Exalead is "a next-generation system," and it warrants a long, hard look by an organization wanting to deliver robust search and retrieval, content processing, and low-latency performance to a single user or tens of thousands of users. You can contact Exalead here.

Stephen E. Arnold, February 25, 2008