eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Theres been a lot of activity at SAP over the past couple of years as it ramps its Enterprise Services Architecture roadmap. But the company is, in a sense, taking a step back so that customers can catch up.
At its recent TechEd conference Sept. 12-15 in Las Vegas, SAP essentially announced that it is revamping its entire release schedule.
Scrapped are plans to develop mySAP ERP (enterprise resource planning) suites to follow the latest version, mySAP ERP 2005 released in June.
Instead, SAP plans to hold ERP 2005 steady for the next five years, releasing any new services, composites applications or functional upgrades in enhancement packages. At the same time, SAP is in the midst of a quiet revolution, should its plans for in memory technology take off.
In August Senior Writer Renee Boucher Ferguson broke the story that SAP may indeed have stumbled onto an Oracle database killer with its in memory capability.
Boucher Ferguson sat down with Shai Agassi, president of the product and technology group and a member of SAPs executive board, in Las Vegas to discuss the companys plans with in memory, with ERP 2005 and with SOA.
I wrote about the BI Accelerator previously and the implications of the in-memory technology are that this could be huge. Is this a real Oracle killer on the database side?
By the way, youre the only one that picked that up. You are absolutely right. Because, you see the largest databases in the world are data warehouses, and those get the most complicated kinds of queries that they need to process as fast as possible.
Theyre extremely hard to maintain, and they use the most CPUs. And the way Oracle prices databases, theyre the most expensive. Thats where the big revenue comes from for them. If you take data warehouses, even if its just the 30,000 SAP customers, and you say, you dont need a database, as a matter of fact what were giving you saves a tremendous amount of database CPUs, youre right, you dont pay Oracle. And the issue here is not do you pay Oracle or not—or if were going after Oracle or not—were just introducing a new way to manage massive amounts of data.
But it is going after Oracles fundamental business?
I think there is a fundamental problem in the way weve done data warehouses in the past. We had to pre-imagine almost every query that the user would send. The job went effectively to DBAs to create the aggregates and DBAs had a lot of work to say, this query came in, it took too long, why was it too long, let me understand the query. Its a very, very complex problem.
We just abused the relational model, because the storage engine was the wrong engine to perform these queries. Weve changed it now to a different storage engine. As a result of that, every query that comes in gets immediate, fast execution.
It changes not the role of the database as much as the data warehouse administrator. That is a huge shift. You looked at just the market for the Oracle database. Im looking at how much money is spent on the restoration of performance on data warehouses and human beings. Its bigger than the database.
Next Page: A new way to store data.
A New Way to
Store Data”>
Do you see in memory technology as applicable across the organization, with other types of transactional applications?
Youre seeing the same engine in the Enterprise Search. The core engine is the same engine. The difference is in search I get you all the hay in the haystack. In BI, I take the hay and make a mattress out of it. We use a similar engine in MDM, but theyre used differently in each case because the store is very different. With those characteristics, think of where we can go with this technology.
Where do you see it going?
In the end, and I dont know how far in the future that is, you will see a new way to store data—all types of data, all kinds of usages. Today, were at the top of the pyramid if you want, at the complexity top of the pyramid, of being able to do the most complex type of applications.
Its interesting because the bottom of the pyramid is also being hit at Oracle, with the mySQL world. So doing the simplest of databases is becoming commoditized. The top is being completely revolutionized, the bottom is being commoditized. There in the middle is not a pleasant place to be.
When you were talking about BIA and in-memory during your keynote, the guy behind me said, where do you store the data? Is there a point where you have to have some sort of backup?
There is a backup. But during run time what youre really storing is in memory. The secret, Ill tell you the secret: everybody stores data in rows; every line item BP sells in their gas stations is one line in the database. When you go to the data warehouse, you flip it. Instead of storing row by row, weve flipped it and start looking column by column. Its very similar with movies.
If you think about taking every picture, stand-alone, and storing picture after picture after picture, youd never be able to screen them fast enough to see a movie over the Internet.
But because youre assuming every picture should be pretty similar to the one before it, and to the one after it, all you need is the delta on it so you can compress it.
Do you see a scenario where you wouldnt need a database under SAP applications?
Long term.
How long? Five years? Ten years?
I dont want to speculate on that. Well tell you when were there. There is speculation, but I dont think we need to speculate on that. For the traditional transactional application right now you need a database. Down the road, youre right … But today? No.
Any connections with IBM on developing in memory?
The IBM youre seeing here is the hardware. We worked on the technology on our own and we worked with Intel on packaging it and optimizing the CPUs. Then we basically took a box we designed with Intel and we went to HP and said, can you build this? And they said, sure we can build a box like that.
And then we went to IBM and they said, oh yeah, wed love to do that. Youll see more. Its a reference platform—very similar to what Microsoft does with Intel that then goes to all OEMs who decide if they want to take it to market.
Is there a point where youve outstripped your customers? And whats the window for them to catch up?
One of the things that Ive tried really hard to do is to simplify consumption. And were getting better and better and better in that. That will change customers adoption cycles, from purchasing to implementation to supporting to operating, and if that happens it will free up a lot of resources on our customers site, which will be applied to absorbing the innovation. Youre right. Were making cars. We have to help [customers] have a better freeway.
And mySAP ERP 2005 is the foundation?
Thats the right of passage
How many commitments do you have for upgrades to mySAP ERP?
Contract wise, I think thats what most people dont get because theres so much hype in the industry about yes, adoption/no adoption. Were now at about 4,000 customers that still dont have the contract for mySAP ERP, that havent signed it yet.
Next Page: Does SOA need to be dumbed down?
Does SOA Need to
be Dumbed Down?”>
So there are probably 4,000 leftovers. But the market has bought [the technology]. The only question left is, when do I move? Not am I going to move? And the 4,000 are buying now at a rate of about 1,500 a year, so were looking at within two years pretty much most of these customers are going to have the right to mySAP ERP.
When I talked to users at TechEd about upgrading, every single one of them said they were upgrading because they have to. Do you agree with the perceptions that users are being forced to upgrade?
No. Lets be really accurate about it. We have a strategy that weve documented and shared and everyone knows about it: 5-1-2. It doesnt mean that after 1 and 2 are over, we stop supporting. We support forever.
Is there any possibility that SOA is a fad?
Its here to stay. Thats like saying the Internet is a fad.
I thought your key note was, in a sense, dumbed down a bit from previous keynotes discussing customers blue print SOA. Do you think that is necessary to get the message across?
Yes. To me, when were able to dumb it down were successful. When you just start with something you still dont get all of it; you dont get all of its meaning. You havent tried the words yet. You havent tried the message. You dont see what customers do. They really tell us what were doing. So when you see us with a very complicated message, were just starting to play around with a concept.
How does SAP differentiate from the pack with its SOA message going forward?
The differentiation will not happen at the level of conceptual discussion. It will happen at the level of execution. And it will happen at the level of adoption. We all say the same thing: SOA is the way to go.
Thats great news. The fact that were not arguing about that tells me that weve dumbed it down to a level that we all agree on. What were saying is a bit different. Oracle is going to start asking: Do you want functionality, or do you want SOA?
You cant take their existing functionality and make that SOA. Youve got the three big ones—Oracle, PeopleSoft and Siebel. Two of them are somewhat close in the way theyve been constructed. Theyre sort of Roman languages, but theyre different Roman languages—you cant go from one to the other. So theyre going to go to one of them: E-Business Suite.
The problem is Siebel was built as an application written into the client, not on the server. You cant service enable a client application, it just doesnt happen. [Theyre] going to have to rewrite it from scratch. To rewrite Siebel from scratch, you dont do it in a year because otherwise the people at Siebel have been really, really stupid.
So to get a service-oriented CRM youre going to get a dumbed-down, rewritten something from scratch. The same thing, by the way, will happen with PeopleSoft because even though they tell you everything was written in Java, the code was written in PL SQL.
Take all that and youre going to get a SOA story that contradicts the functionality story. We have all the functionality, all service enabled. Our number is 05, not 08 maybe. And were not half way there; were next gening.
And Microsoft?
Microsoft is very close to what Oracle has. Dynamics is a brand, but Axapta, Navision, Great Plains, again different companies, different dictionaries. Some of them are Danish, some are in Fargo. I dont know if you know, by the way, that Doug [Burgum, head of Microsoft Business Applications] is gone today.
What do you think of Satya Nadalla [Burgums replacement]?
Hes an R&D guy, one level up. It wasnt our R&D guy that was the visionary in the group, he was developing. The almost guru-like level that Doug played in that ecosystem, looking at midmarket and looking at all the ecosystem [with MBS], I would be very nervous.
Its not that they picked up a great executive and plugged him in there. They grew up a guy who is head of development. Im questioning that whole thing. But thats their life. What Im trying to put in perspective is were rolling. When this game is played if you get 10,000-plus ERP implementations with thousands of services on them, with hundreds of thousands of ISVs with solutions around them—it doesnt matter what they ship.