Exclusive Interview: Satish Gannu, Cisco Systems Inc.

August 24, 2010

I made my way to San Jose, California, to find out about Cisco Systems and its rich media initiatives. Once I located Cisco Way, the company’s influence in the heart of Silicon Valley, I knew I would be able to connect with Satish Gannu,  a director of engineering in Cisco’s Media Experience and Analytics Business Unit.  Mr. Gannu leads the development team responsible for Cisco Pulse, a method for harnessing the collective expertise of an organization’s workforce. The idea is to apply next generation technology to the work place in order to make it quick and easy for employees to find the people and information they need to get their work done “in an instant.”

I had heard that Mr. Gannu is exploring the impact of video proliferation in the enterprise. Rich media require industrial-strength, smart network devices and software, both business sectors in which Cisco is one of the world’s leading vendors. I met with Mr. Gannu is Cisco Building 17 Cafeteria (appropriate because Mr. Gannu has worked at Cisco for 17 years). Before tackling rich media, he served as Director of Engineering in Cisco’s Security Technology Group. I did some poking around with my Overflight intelligence system and picked up signals that he is responsible for media transcoding, a technology that can bring some vendors’ network devices to their knees. Cisco’s high performance systems handle rich media. Mr. Gannu spearheads Cisco’s search and speech-to-text activities. He is giving a spotlight presentation at the October 7-8, 2010, Lucene Revolution Conference in Boston, Massachusetts. The conference is sponsored by Lucid Imagination.

cisco satish gannu

Satish Gannu, Director of Engineering, Cisco Systems Inc.

The full text of my interview with Mr. Gannu appears below:

Thanks for taking the time to talk with me?

No problem.

I think of Cisco as a vendor of sophisticated networking and infrastructure systems and software? Why is Cisco interested in search?

We set off to do the Pulse project in order to turn people’s communications in to a mechanism for finding the right people in your company. For finding people, we asked how do people communicate what they know?  People communicate what they know through documents — web page, or an email, or a Word document, or a PDF, and now, Video. Video is big for Cisco

Videos are difficult to consume or even find. The question we wanted to answer was, “Could we build a business-savvy recommendation engine. We wanted to develop a way to learn from user behavior and then recommend videos to people, not just in an organization but in other settings as well. We wanted to make videos more available for people to consume. Video is the next big thing in digital information, from You Tube coming to enterprise world.  In many ways, video represents a paradigm shift. Video content takes a a lot of storage space. We think that video is also difficult to consume, difficult to find. In search, we’ve always worked from document-based view. We are now expanding the idea of a document from text to rich media. We want to make video findable, browseable, and searchable. Obviously the network infrastructure must be up to the task. So rich media is a total indexing and search challenge.

Is there a publicly-accessible source of information about Cisco’s Pulse project?

Yes. I will email you the link and you may insert it in this interview. [Click here for the Pulse information.]

No problem. Are you using open source search technology at Cisco.

Yes, we believe a lot in the wisdom of the crowds. The idea that a community and some of the best minds can work together to develop and enhance search technology is appealing to us. We also like the principle that we should not invent something that is already available.

I know you acquired Jabber. Is it open source?

Yes, in late 2008 we purchased Cisco bought the company called Jabber. The engineers had developed a presence and messaging protocol and software. Cisco is also active in the Open Social Platform.

Would you briefly describe Open Social?

Sure. “Open Social” is a platform with a set of APIs developed by a community of social networking developers and vendors to structure and expose social data over the network, at opensocial.org. We’ve adopted Open Social to expose the social data interfaces in our product for use by our customers, leveraging both the standardization and the innovation of this process to make corporate data available within organizations in a predictable, easy-to use platform.

Why are you interested in Lucene/Solr?

We talked to multiple companies, and we decided that Lucene and Solr were the best search options. As I said, we didn’t want to reinvent the wheel.  We looked at available Lucene builds. We read the books. Then we started working with Lucid. Our hands on testing actually validated the software. We learned how mature it is. The road map for things which are coming up was important to us.

What do you mean?

Well, we had some specific ideas in mind. For example, we wanted to do certain extensions on top of basic Lucene. With the road map, open source gives us an an opportunity to do our own intellectual property on the top of Lucene/Solr.

Like video?

Yes, but I don’t want to get into too much detail. Lucene for video search is different.  With rich media sources we worry about how transcribe it, and then we have to get into how the system can implement relevancy and things like that.

One assumption we made is how people speak at a rate of two to three words per second.  So when we were doing tagging, we could calculate the length of the transcript and size of the document.

That’s helpful. What are the primary benefits of using Lucene/Solr?

One of our particular interests is figuring out how we can make it easy for people in an organization to find a person with expertise or information in a particular field. At Cisco, then, how our systems help users find people with specific expertise is core to our product.

So open source gives us the advantage of understanding what the software is doing. Then we can build on top of those capabilities., That’s how we determine what, which one to choose for.

Does the Lucene/Solr community provide useful developments?

Yes, that’s the wisdom of the crowds. In fact, the community is one of the reasons open source is thriving. In my opinion, the community is a big positive for us. In our group, we use open social too.  At Cisco, we are part of the enterprise Open Social consortium, and we play an active role in it.  We also publish an open source API.

I encourage my team be active participants in that and contribute. Many at Cisco are contributing certain extensions. We have added these on top of open social. We are giving our perspective to the community from our Pulse learnings. We are doing the same type of things for for Lucene/Solr.

My view is that if useful open source code is out there, everyone can make the best utilization of it.  And if a developer is using open source, there is the opportunity for making some enhancement on top of the existing code. It is possible to create your own intellectual property around open source too.

How has Lucid Imagination contributed to your success in working with Solr/Lucene?

We are not Lucene experts. We needed to know whether it’s possible, not possible, what are the caveats. The insight, which we got from consulting with Lucid Imagination helped open our eyes to the possibilities. That clinical knowledge is essential.

What have you learned about open source?

That’s a good question. Open source doesn’t always come for free.  We need to keep that in mind. One can get open source software. Like other software, one needs to maintain it and keep it up to date.

Where’s Lucid fit in?

Without Lucid We would have to send an email to the community, and wait for somebody to respond. Now I ping Lucid.

Can you give me an example?

Of course. If I have 20,000 users, I can have 100 million terms in one shard. If I need to scale this to 100,000 users and put up five shards, how do I handle these shards so that each is localized? What is the method for determining relevancy of hits in a result set? I get technical input from Lucid on these types of issues.

When someone asks you why you don’t use a commercial search solution, what do you tell them?

I get this question a lot. In my opinion, the commercial search systems are often in a black box. We occasionally want to have use this type of system. In fact, we do have a couple of other related products which use commercial search technologies.

But for us, analysis of context is the core. Context is what the search is about. And when you look at the code, we realized, how we use this functionality is central to our work. How we find people is one example of what we need. We need an open system. For a central function, the code cannot be a black box. Open source meets our need.

Thank you. How can a reader contact you?

My email is sgannu at cisco dot com.

Stephen E Arnold, August 24, 2010

Sponsored post

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta