SAP executive board member Shai Agassi has positioned the company in a leadership role in the emerging event-based, real-time architecture. In a conversation with eWEEKs Steve Gillmor, Agassi talks about SAPs big bet on NetWeaver and its Enterprise Services Architecture.
The last time we talked a year and a half ago, there were some doubts—not by you, but others—about Web services and the impact of what is now known as Service-Oriented Architecture (SOA). When are you shipping?
This year. Its here. One of the analysts told us, When you announced this a year and a half ago, we said its a nice strategy, a nice set of slides. And the reality is we actually built the thing. Most other people just announced slides, you guys actually built a thing. I can hold it.
Now, youve got a thing, and its commoditizing the business process layer …
I wouldnt say its commoditizing the business process layer. It productizes the business integration layer. Thats one of the big mistakes that people made who looked at what Web services will do—jumped too quickly to the conclusion that the business process layer will be commoditized.
I dont believe customers will go out and buy a business process or a single Web service. What customers will buy are solutions, but what they want in solutions is flexibility.
And thats where Web services will change the future of business. They will still buy a solution to a business problem, but because it is being built with Web services, they will be able to interchange the pieces—and usually they will interchange 1 or 2 percent. Its like cars—you dont go out into the aftermarket and change the whole car. You change the radio, you put some things on the roof, but…
You custom-build it.
How many people custom-build cars?
You can get a Ford car custom-built for you five days later.
What did it do? It changed the business model for Ford. It gives Ford the ability to become a better supplier for their consumer. What does this do for us? It gives us the exact same thing that you describe with Ford. It allows us to build almost custom solutions for the customer—not a one-off, but a one-industry-off, if you will.
Vertical slices of horizontal applications.
If you take your xApps as packaged …
As the end result of your architecture and intellectual property …
Service enablement of the core applications.
A Track that Can
Factoring in the IT doesnt matter discussion—commoditizing and eliminating the differentiation between technology solutions, leveling the playing field—what are you going to provide by building on top of the xApps architecture —
The Enterprise Services Architecture (ESA)
—To give individual companies yet another competitive advantage on top of this new architecture?
In IT doesnt matter, there was a comparison between the IT and the railroad business, where we got into the Gold Rush of railroads and overbuilt infrastructure. At some point you didnt need any more tracks, and the railroad business disappeared.
The one thing they forgot to mention is that the IT business—especially an enterprise-applications business—reflects the business processes and the best practices in the business. And those keep changing all the while. Its as if you have a train, but it takes a different route every year or so.
Thats the big shift that is happening. We have to assume that the business will change every year, two years, three years, and we have to build a track that knows how to move itself. Were in the transportation business.
The first thing we said is, Give me a platform where I can build things—and reconnect them on the fly—as I go through my new strategy, my new plan. Second, we went to the applications guys and said, Service-enable your pieces so I can actually connect them easier. I dont need to service-enable every time, I want you to do it once. And yes, I know its hard and I know its expensive, but youve got 20,000 customers who are paying for it through their maintenance.
And then we said, What can I innovate as a result of that, what can we build new that we couldnt do before, which kind of processes can we do? Theres a whole family of end users and roles in the business that are almost ignored. And were targeting them now with this innovation.
We looked at the guys who did the repeatable, mundane, transactional business on a day-to-day basis, the guy who puts the data in financials into the GL [General Ledger], the girl who puts the accounts receivable into the accounts payables. What about the strategic knowledge workers in financials? The guys who do analysis of which companies to buy?
The Mergers and Acquisitions module.
Right, the post-merger integration. We dealt with the guy who puts the build material into the PLM [Product Life-cycle Management] product. What about the guy who deals with what the innovation looks like? The guys who invent the next product? The guys we dealt with were the $30,000, $50,000 a year employees. And we helped them become 10, 20 percent more efficient, which is how we are seeing today companies grow their revenue without adding more head count.
So, youre responsible for the jobless recovery.
Part of the problem is that the recovery weve seen so far is an efficiency recovery. Youll see another wave of recovery coming from innovation and growth.
Given this packaged middleware layer …
We call them packaged horizontal scenarios. Instead of process scenarios, these are horizontal scenarios—sharing information, making decisions, managing master data across the enterprise.
Say youve got two companies [or supply chains] adopting this technology. Again, how do you provide differentiation, competitive advantage, between those two?
At the end of the day, we can build hundreds of new processes that we couldnt do before, leveraging the intellectual property that we have in the services modules and underlying components.
Does that mean you sales-rep the xApps?
Either coming as xApps or coming as industry solutions or coming as horizontal solutions. Were doing an xApp for product innovation—the management of the pipeline of ideas in the company. Almost like youre managing opportunities in sales—managing the pipeline of ideas all the way from the first incarnation of the idea all the way out to the shipping of the product and launching it into the market. Managing the life cycle of projects and allocation of skills into the project—research, program management, managing your portfolio—are key things that apply to every single company in the world.
Tackling Time Management
Are you familiar with RSS?
The company that can aggregate Weblogs and combine them based on topics instead of based on the person?
No, that would be a Bloglines or Technorati. Im really talking about this emerging identity-based platform and real-time infrastructure. If you look at Google, for example, its essentially providing useful results based on the dynamics of page rank, a trusted authoritative reputation derived from domain expertise. Youve talked about an alert mechanism youre developing…
The proactive business event network.
That seems to be the true platform.
I think thats probably about two or three years away, but that my vision—the goal of what were doing. Were shifting from different drivers for why the businesses are building their fabric—the infrastructure. Whereas before it was about management of resources, then it shifted to management of information, [now] were shifting into management of time.
The biggest driver is time to change. It could be a small event: Somebody writes me a note, and I need to respond to that note. A customer is calling in and their product is not there. Or an XML message comes in that says, Im doing an RFP, and I have 30 minutes to respond to a TI or an AMD. Can you respond to this in 30 minutes?
Its beyond identity management. Its now: Who deals with this event, the systems and the people? And the relationships exist already in the information space. They are the drivers for finding these pieces of information. The first thing I do… Is that product in inventory? Its an exception. Who deals with this exception? Whats the relationship between this product and this customer and the people that I have in my business, and how do I drive that event to resolution within 28 minutes and counting.
And you look at these short events—they happen so often in business. Were dealing with millions of events a day. They propagate and they will create more and more events—and were dealing with it in e-mail.
Which is …
Ninety percent of all collaboration applications are built on an e-mail infrastructure by default.
The worst thing about it: Its not predictable. You may succeed seven out of 10 times, but the other three, you lose the deal. Now, that time fear is going to start driving companies on the small events, but on the long events—a merger, a project, a product life cycle—the time to change from one design to another design, from one company to the merger of the two companies—its about managing that time, being able to shrink that down.
Whirlpool—they can finish a design and the R&D of a product, and it takes them a year to get the product on the market. You know what happens during that year? All the Koreans get to the market before them. Why? Because they copy their designs—they see them in a show, on the floor, and they copy them—and they have a three-month time to market.
Same thing happened to Philips. Philips invented all the key innovations in the consumer electronics space, and they always were three months behind Sony. Even on stuff they invented, DVDs, they came in after. Manage that time to change, and you become Dell. You dont manage it, you become dealt.
Strategically, how do you move forward in this space?
The fundamental engine underneath this I call the business event network. Its the fundamental engine that drives events at all layers, propagates them and aggregates them.
The Protected Garden
Are they XML-based?
All of them are XML-based, all the time. If you dont think of them as XML-based—[if] you think about it in a proprietary format—only you can understand the messages, which means you cant tie the ecosystems.
Inside the business, once the event comes in as XML, you can optimize it and do WSRM [Web Services Reliable Messaging]. You can say, Since I know that Im talking to myself, to my own system, I can byte-compress it, I can do all kinds of things to improve performance. But the minute I cross out of the protected garden, its always XML.
Microsoft has always done this—across a VPN, between satellite offices—theyll use their own protocols for performance. As you develop your network of endpoints with xApps and the ESA, are you building what you once called a high-speed expressway?
The Boston Big Dig? Only in places where were digging inside our own systems.
Along a supply chain in a service-oriented architecture, your own system would include partners and subcontractors.
Always XML, because when I go outside my company, I dont know whats on the other side. Now, if the two of them can actually do a handshake and say, Hey, by the way, we can optimize this, great. I dont believe that outside the company, optimization is so big. When I look outside, the big latency is network latency. Compressing, decompressing XML is not going to make a huge difference.
Inside the company, I can build a huge pipe between my business processes written in ABAP, which I cant change—it runs my business, its been tested, I cant touch it—and my Java infrastructure so I can do my customization. That pipeline, which is getting crossed a hundred thousand times every second, can be optimized. Excellent—you gave me something that nobody else can do. Thats what we can dig underneath our own town to do this. But when I go outside the company, or between divisions, there is no value in changing this.
How do you capture the value of the ESA at a time when were moving to this real-time platform, with mobile devices, etc?
Youre right on the mark. The shift is not just in which device Im using to read it, it is whether the framework is proactive or reactive. Do you carry a BlackBerry?
The big difference in the Berry is not its screen, nor the keyboard, nor the semi-connectedness. The difference is the push model that it deploys. You always assume that its there—why? Because you didnt see the event until you had enough context about the events. The same thing happens here—the network needs to work behind the scenes seamlessly to actually expose to you just the events you need to know. I think thats the one thing we all love about our BlackBerries: They can filter the viruses out…
Certainly its the lowest common denominator that filters them out.
Were going into a model where the business network will become proactive. When I get the event, I wont get it just to you, I get it to everybody who needs to deal with it. Everybodys on the same page. And they get it within a context. Where is the context coming from? Mostly from master data.
To use the RSS terminology, if I subscribe to a certain number of XML information feeds and put them in a certain order, an attention list—whats most important for me to listen to first—those kinds of decisions are the ones…
Thats what the event processor does.
How ViewPoint Works
Taking that as a black box and distributing it across the supply chain seems to be a job for somebody. Is that your job?
Which is what were doing. We have a piece of technology built on NetWeaver—what we call ViewPoint. We havent announced it, we arent talking about it too much—but what Viewpoint allows us to do is to take a lot of these feeds and, at the object level, assign importance.
What it allows you to do is get all the feeds from all the systems and, as you read through them, highlights words that it recognizes as objects. Enterprise portal, business intelligence, whatever it is that youre looking at. And all you have to do is hover over the words and say, This is more important to me or less important to me.
And what it knows how to do now is create almost like a news.google.com just for you from all these feeds—deliver just for you what youre supposed to see and allowed to see—in the priorities which you would like to see. Not only that, it allows you to choose also how you want it to be presented, so it knows which device youre reading it on, how much you can read at each point in time, and sets the page for you.
Once youve read the event and responded to it, and you can respond to it in various different ways, you can just say ignore it from this point on. Clicking on it wipes it out. The next time you see the page, its gone—Ive dealt with it. Or you can say, this is so important Im going to start a discussion group around it—and invite people to my discussion group around this.
You can create a blog effect around an object or an article or a person—the writer, the author—because theyre all objects. You said, Im only measuring based on the author, thats the only metadata you can measure on. Imagine you could say, I want to look at blogs based on these objects—if theyre present in there, you see Shai Agassi, NetWeaver, ESA—interesting. Even if the author is somebody Ive never heard of.
The reputation of the person who said that is part of the weighting of the algorithm.
Once you get the article, you say this guys great, I love what he said. Now, youve got multiple levels of looking at this stream at the object level, a true understanding. And you might say, whenever a new thing comes from this person, not only put it in my bucket, alert me—but send me an instant message that theres a new thing to read.
How do you handle instant messaging in this environment?
Its one of the features of our collaboration piece. Were trying out other instant messengers, but were doing it again within the context of the system that we have.
Are you capturing this traffic?
We capture the instant messages, but more importantly, we can create virtual presence capability. You can say, I want to start an instant message not with Steve, but with a person I dont know. Why do I want to start it? Because I have this event: I want to find the project manager for this product that is mentioned in the event, and I want to start an instant message with that guy. Find that relationship, start an instant message with a virtual entity.
In this fabric of information objects that are floating around, this is an uncaptured realm.
Yes and no. We take a lot of these event resolutions and we capture them, not for the sake of recording the history, but for the sake of trying to find repetitive patterns in these resolutions. What we do from these repetitive patterns is create something called guided procedures.
The next time an event comes in, [the system] says, Here are the seven guided procedures that we found for resolving this event. You have an RFP—guided procedure No. 1: If you dont have enough time, find the customer representative, call the customer and say, Can I get another hour? And if you say, Yes, thats what I want to do, it takes you through step one, step two, step three, and if youve got another hour, everybodys updated that Ive got another hour to respond.
Another guided procedure is that if you cant find the customer representative, go find if there are alternate suppliers for this component. If we dont have it in inventory, maybe we can order it and guarantee delivery. Proposal No. 3: Go to the partners and see if they have inventory. And these guided procedures come not out of thin air—they come out of monitoring what weve done in order to resolve it in ad-hoc scenarios, and taking some of these ad-hoc scenarios, saying this is a great idea.
Hub and Spoke
Its striking that youre not aware of RSS.
Believe me, Im going to Google it the minute were done.
Microsoft has difficulty playing outside their desktop sandbox. Thats why I keep coming back to the relationship between the nodes of this network fabric that youre building.
Youre hitting on an interesting point. The big question is: What becomes the hub and what becomes the spoke? One of the models that could emerge is that the hub would become your personal device that you carry on your hip, and the spoke would be any desktop device you get close to—becomes your extension of your hub.
The chip inside this [Nokia] phone is in fact a Java card. But these guys are not yet capturing my identity in a meaningful way.
They have a lot of identity, but they do not become the identity. It may not actually be that device, even though Nokia is probably one of the most advanced in that arena. But at some point in time, that may be your one and only device, one and only identifier.
And potentially it alerts …
Some more high-bandwidth-connected device and says, OK, I take over.
How does that plug into what youre doing?
Because these events propagate to you not when youre next to your desk. These events propagate to you when youre everywhere, and what you need is a device that can help you resolve the event, maybe getting less context or less ability to display the context when youre looking at it on this type of device.
But if its a big event, you dont want to go and find a machine, figure out how to install something on a machine, and then…. You want to go to the small machine in the corner of the room, put your phone next to it and say, Let me deal with this thing—sorry, Ive got some emergency, or let me run to any hub around it. Now, are we there? No. Are we going to be there? Yes.
Theres a unification of identity across the personal and professional space. You only have a certain amount of time.
There is a certain amount of identity in the systems themselves. Systems will talk to one another. Context will be distributed. Yet at the end of the day, you have to ask yourself: Who can give me the context?
When an event comes to me, I get an e-mail. Name a company—Whirlpool. Whirlpool wants to buy more software from us. Theyve asked about the following products. They have the following different stuff in their shop. They have a time limit of X time, and here is the estimated budget that we think they have.
When you get this e-mail, if I could get around it immediately—Who are the people at Whirlpool that weve met with? What have been the conversations? What have they bought already? Whats their tendency—do they buy quickly or slowly?
All these things—where you find them—thats your data. Its the master data and the analytics around master data. Now, do you find that in the technology platform or the application platform? My claim is that the master data management—the brain of the business—is in the applications today. And you cant take that out of the application and serve it from somewhere else because its no longer the truth. Its a version of the truth.
Its a snapshot.
And its not true anymore. The minute that you take it out, its no longer true. You cant have the person and the brain in two different places. It has to be together. Who has the ability to actually create a true brain for the applications, to have enough gravity to serve master data to the whole organization?
What about a Salesforce.com?
Salesforce.com is a really interesting question, but its a derivative of its infrastructure. If I have a big infrastructure that allows me to serve these things, do I need Salesforce.com? Big question. Can Salesforce.com give me the fluidity and transparency and context that kind of a business event network can give me? No, they give me just this one aspect—my view of the sales pipeline—nothing else in the business. When I get an event, I have zero information, zero ability to help.