SAP's Agassi Fields Queries, Zings Oracle

He was in Las Vegas to unveil updates to SAP AG's projects with Ariba Inc., Intel Corp., Microsoft Corp., Oracle Corp., and others.

Agassi, president of SAP's product and technology group and a member of its executive board, also sat down with reporters, including TechWeb's Laurie Sullivan, to explain the NetWeaver platform, composite applications projects, database commoditization, and the $1-billion-per-quarter for services the German software company expects to generate.

TW: What percentage of SAP customers use NetWeaver?

Agassi: We have 31,000 NetWeaver components installed. Depending on the layer, at some layers such as master data management, we are in the very early stages. On others such as analytics, we are in the middle of mainstream adoption with more than 10,000 installations. If you look at NetWeaver projects, for example, many layers many processes, we are looking in the next 12 months at between 5,000 and 6,000 customer implementations with multiple installations. We're hoping it doesn't go beyond that because you don't want it to become an uncontrolled growth. Most customers will start between January and May. To put it in context, 6,000 installations is PeopleSoft. To explain the magnitude of NetWeaver adoption, we're doing sort of a PeopleSoft year.

For the services market, we're estimating that between 30,000 and 40,000 will need to be deployed in the market. You're looking at a more than $1-billion-per-quarter service industry. We're being conservative saying [that]. Where have you seen a service industry grow to a billion dollars a quarter in two years from zero? NetWeaver was launched in 2004.

TW: SAP is introducing many composite applications. Do you see SAP going beyond enterprise applications and into operating systems?

Agassi: We won't do Xboxes. We're an applications company at the core and at the heart. We've never seen a reason to move to commoditized layers like operating systems. Unless we can create tremendous differentiation for the overall stack, veering off to try and take over the operating system market doesn't make sense. We went to middleware because we've always been in middleware. Throughout our history we've built middleware for our own products. We've built abstract operating systems, databases, communications, and put more application-servers installations in the market than any of the application server vendors--we just never sold them. We gave them as part of our application. When we built the next generation of middleware, NetWeaver, we needed to make a decision on whether to keep it inside, hidden where no one sees it, or put it outside. And it became obvious that the fact we didn't sell middleware complicated our customers' lives because they couldn't take what we built and build on top of it other solutions.

TW: SAP collaborates with Ariba, Microsoft, and Intel. Do you see one day collaborating with Oracle?

Agassi: We are Oracle's largest customer. And I have to tell you it's very difficult to collaborate with them. [In his closing remarks to analysts, Agassi addressed the war with rival software titan Oracle and said, it's "a war we didn't ask for and didn't declare."]

TW: Competitively speaking, what are your thoughts about the database layer and is it something that eventually has to be dealt with?

Agassi: The database layer as we know it will go through radical shifts. Two major shifts are happening in the market. One is service-orientated architecture. We're raising the level of abstraction one level up. As a result you can't move level up to a conversational level of objects, properties and services, and still need to know how the table structure is in the database underneath. Once you move up to service orientation the importance of the database and structure and its capabilities diminishes by orders of magnitude. The second things we're seeing is a fragmentation of the big-beast database approach into multiple segmented special-purpose data stores. We are seeing it with master data management that has special characteristics. We know the indexes and all the data needs to be in memory at all times and the properties around it need to be outside the store, not inside, so I can plug other stores into it. It's a different approach than transactional where data and indexes are on disk. I can cache out pages and put my code inside the database.

Data stores are diametrically opposed. Between them we have high-performance analytics, the engine of our analytics solution where the indexes are in memory, the data on disk and you have no code. You look at these approaches and they are fit for a special need. And yet what you want is a single unified repository on top of that. The enterprise services repository that says when you need to get that property out of that object you go to this store. Oracle is trying to say that everything is in a unified single database. We say it doesn't give you performance. It's the meta-data that matters more than the data. You'll see that transition happening during the next four to five years. Database will matter less and become more commoditized, and Moore's law will push the guys from the bottom to form everything you need for 99 percent of the companies.

TW: Open-source vendors are making noise about moving into the enterprise market by becoming certified as databases become commoditized. Are they ready?

Agassi: All are getting ready, including Ingress and MySQL. We work with MySQL to get them ready. There are more fundamental issues. They are pushing on the Sequel Server guys. With all the functionality you're seeing for the first time Sequel Server for free. That pushes on the IBM guys and puts DB2 Express for free in the market. That pushes on Oracle, which put an Oracle Express for free in the market. None of these guys are doing it because they like to give away software. At some point the price erosion at the bottom will happen. It should have happened earlier. With a single unified standard of access, SQL's been in the market for 15 years. At some point you are bound to get many vendors doing the same thing. You can't differentiate on any technology standard for that long. We have to admire Oracle for keeping the database business floating for a long time. Unfortunately, it's not grown any other business for them.

TW: Which database vendors is SAP working with?

Agassi: We are working with MySQL, Oracle, DB2 and Microsoft on the MySAP layer. We are working with Sybase on multiple areas, but you have to realize there is tremendous complexity in introducing a new database into the SAP ecosystem to run enterprises. It's not just completing certification. It's testing, support centers, and a whole set of events.

TW: Where do you see the general investment climate and is there a shift?

Agassi: The shifts are more at the business level than IT. The IT industry seems to think we are the dog wagging the tail. We are the tail. Especially in North America, more companies are finding the curve of product innovation has run its course. When you can't differentiate your product inherently, you have to move into process innovation. I love Starbucks. Most people say they typically ask for one specific drink whenever they go. This is because they get the same delivery of that coffee everywhere in the world. This is not a product issue. It is a process issue. It's a mechanism that allows you to train tens of thousands on a new drink everywhere around the world. That process of delivering the knowledge to employees allows Starbucks to charge $4 for a cup of coffee in Shanghai. Next door there is a local Chinese espresso maker. I don't know if it's better or worse, because I didn't taste it, but I know its 40 cents. The process differentiation creates a brand and 90% margin.

You see other companies doing this too. Wal-Mart is not differentiating on brand. They differentiate on the process of sourcing goods in China and sending information from their shelf to suppliers. The Home Depot has fantastic ideas on how they will differentiate themselves. They're not differencing on IT systems, but rather on business processes. Companies are putting more financial investments into process innovation than research and development because the only tools that support process innovation is a process backbone. That is the shift to service-oriented architecture.