Saturday, April 14, 2012

SAP’s Database Strategy Faces an Uphill Battle

Last week I attended a half day SAP press conference in San Francisco, on the subject of SAP’s strategy for database technology and mobility. Both are hot topics in enterprise software, and there were plenty of announcements. In fact, it’s quite easy to get lost in the weeds. So, in this first of two posts, I’ll try to summarize what I see as the big picture for SAP’s database strategy.

SAP Positioning Itself as a Database Company

When SAP acquired Sybase in 2010, it said it was doing the deal primarily Sybases's mobility platform. But Sybase also has its traditional relational database products, leading with its Adaptive Server Enterprise (ASE) database. At the same time, SAP itself has been rolling out its HANA in-memory database (IMDB) technology. Until now, these two database products were managed separated, but no longer. SAP is consolidating all its database offerings—HANA and Sybase’s—along with middleware and tools, under one management unit.

There were many tactical announcements. SAP announced general availability of its BW business intelligence product on HANA, and its plan to make HANA available later this year as the database of choice for its small business customers of its Business One ERP product. In addition, customers later this month will have the flagship Sybase ASE database available as a deployment option for SAP’s Business Suite and All-in-One products.

SAP rolled out all of these announcements under the banner of its plan to become known as a database company.

Database Migrations Difficult to Justify

After the press conference, one SAP executive sensed my ambivalence about this plan. With Oracle taking an ever-increasing adversarial position toward SAP, I can understand SAP’s discomfort with having a large percentage of its best customers running on Oracle’s database. At the same time, the other two major providers of relational databases (IBM and Microsoft) are SAP-friendly. IBM is SAP’s largest system integration partner, while SAP and Microsoft often find their technology interests aligned. So, how do you threaten Oracle while not also threatening IBM and Microsoft?

Furthermore, does SAP honestly believe that existing SAP customers are going to migrate in droves from Oracle, IBM’s DB2, or Microsoft SQL Server to HANA or ASE? In the case of business analytics, there may be some movement toward HANA, yes, as the value of in-memory performance for analytic applications is somewhat easy to envision. But what about SAP’s business applications, such as Business One, All-in-One, and the Business Suite? With all the challenges and demands placed on CIOs these days, it’s difficult to imagine an installed SAP customer undergoing a database migration, simply to eliminate some Oracle, or DB2, or Microsoft SQL Server licenses. SAP insists there is business value for HANA in some transaction processing—and I can see that, say, in supply chain management. But is that enough to justify a database migration? Even less so, why would a customer swap out Oracle, DB2, or Microsoft for ASE, which is essentially a like-for-like product? I just don’t see it.

In side-bar discussions, SAP executives basically agree. Alright then, so the target is net-new application customers? But here the challenge is essentially the same. In most cases, business apps prospects already have skills and experience with Oracle, DB2, or MS SQL Server. Are they really going to want to invest in learning Sybase ASE, or HANA? Unless they can completely eliminate those other database platforms from their environments, going with Sybase or HANA is adding to their complexity, not simplifying things.

I think that selling databases is going to be harder row to hoe than SAP is making it out to be.

Subsidizing HANA May Meet Complications

Perhaps recognizing the challenge, SAP realizes it is going to have to sweeten the pot, especially for HANA. So, at the press conference, SAP announced that it is putting up some serious money, through two funds:
  • For new and existing customers: a $337 million fund to subsidize services for SAP customers to convert to HANA. I assume the initial target for these funds will be in migrating business analytics customers to HANA.

  • For technology start-ups: a $155 million venture capital fund through SAP Ventures for start-ups to build new apps on HANA.
It’s encouraging that SAP is putting its money where its mouth is. For customers, making a database change may not be cost-justified without some help from SAP. Moreover, tech start-ups may need some financial encouragement to build new products on HANA.

However, I see complications with each of these funding efforts.
  • With the customer fund, there may be issues with SAP’s partners. By funding SAP’s own services to assist with HANA, SAP is taking work away from partners, who typically play a key role in SAP implementations and migrations. In response to my question on this, SAP executives said that it will bring partners into this work at some point in the future.

    Nevertheless, I have to believe that, at first, partners will view SAP as increasing its share of services at the partners’ expense. This is especially true under current economic conditions where customers can only absorb a certain amount of change at once. Moreover, by delivering HANA services directly, SAP delays giving partners the HANA experience they will need for the future. SAP can solve this problem, of course, by ponying up the money but letting customers choose whether to use SAP’s professional services group or partners to deliver the services, or by co-delivering services with partners.

  • For start-ups, HANA may not be as attractive as SAP thinks. Looking back at SaaS and other tech start-ups over the past decade, most of them chose to build on open-source database technologies, such as MySQL or PostgreSQL. The reason, of course, is that open-source infrastructure minimizes their own costs as they grow. It also leaves more customer budget available to invest in the application, instead of the required infrastructure.

    I once asked a start-up executive why his firm was building on MySQL instead of Oracle. He replied, “Oracle scales technically, but it doesn’t scale economically.” I have to wonder if HANA will face the same resistance, even with funding from SAP Ventures. A quick check with associates indicated that there are already open-source in-memory databases (IMDBs), including CSQL and VoltDB. I have no knowledge of the capabilities of these products or how they compare with HANA. It is likely that HANA is head and shoulders above open source alternatives. But Oracle’s flagship database was and still is head and shoulders above open source capabilities, and that didn’t stop cloud start-ups from using MySQL and PostgreSQL.
Ultimately, very few organizations want to buy databases—or middleware. They want business applications, and those apps require databases and middleware as part of the technology stack. So, when SAP talks about becoming a database company, it’s hard for me to become excited.

Perhaps SAP already knows that it's going to be difficult. Earlier this year, it began floating the idea of making its goal, "to become the No. 2 database provider by the year 2015." But by the time of the press conference, the goal had been watered down to “becoming the fastest growing database provider.”

When you are starting from such a small market share, becoming the “fastest-growing” is not a very high bar.

Update, Apr. 16: Some deeper questions on Oracle's database strategy from Jonathan Wilson. And, a good post from Vitaliy Rudnytskiy, pointing out that HANA is more than an "in-memory database."

Related posts

SAP in Transition on Mobile, Cloud, and In-memory Computing
SAP Innovating with Cloud, Mobile and In-memory Computing


Vital BI said...

It is a bit sad to observe, that indeed some SAP folks cannot articulate their advantages in right words. HANA's value and innovative idea (yet still to be seen) is in the ultimate promise of combining OLTP and OLAP on the single platform ( Yet, in my opinion they were quite successful in keeping customers and the market hot about HANA for almost 2 years now.
Otherwise, all your comments are extremely right. Especially about upsetting SIs, including me whose day job has been SAP analytics consulting for almost 10 years now.

Frank Scavo said...

Thanks for the feedback, Vitaliy, and I encourage everyone to read the post you linked to above.

Ranjan said...

it's the ERP that needs overhaul, you cannot expect high performance from ABAP stack which is not concurrent or distributed computing enabled. You also need to design IMDB in lockless manner to reap benefit in OLTP scenario requiring more memory. In all they will have to fight a long, tough, unsettling battle with hardly any real IP with Oracle, where they don't have the price advantage, IMHO IMDB is only suitable for CEP and analytics, otherwise whats the point

Jon said...

I agree with Vitaliy, but the ultimate promise for HANA is more than just combining OLTP and OLAP. It's about combining the multitude of independent repositories (SAP has at least one for each product) into a single repository and eliminating all the integration, duplication, and complexity that current SAP environments have, while at the same time providing a solution that is not only faster, but also more flexible and scalable. Databases might not get CIOs excited, but a platform that truly delivers a single source of truth, underpinning a new breed of enterprise applications, world-class analytics (and maybe even mobilty), that should be enough to get a discussion going.

As for justifying database migrations, for customers running SAP on Oracle this is not as difficult to justify as you might imagine. If you are already considering either a hardware renewal or unicode conversion, then combining this with a DB migration is really not that much additional effort or risk, and the benefits in terms of licence savings can be significant. I work for a SAP Services Partner in Australia that has performed many of these migrations in recent years, with customers moving from Oracle to either DB2 or SQL Server.

In our region we also have quite a few customers who are running MaxDB (a database which SAP has owned and sold for around 15 years - and which seems to have been forgotten in recent discussions), often on Linux. In terms of a cost effective platform, this is pretty hard to beat. I can't see any reason why these customers would migrate to anything apart from a consolidated HANA environment, which delivers functionality that is not available on current DB platforms.

Frank Scavo said...

Thanks, Jonathan. One question, you wrote: "It's about combining the multitude of independent repositories (SAP has at least one for each product) into a single repository and eliminating all the integration, duplication, and complexity that current SAP environments have..."

Are you saying that when SAP allows HANA to underly the Business Suite, that HANA is going to actually change the current Business Suite database schema? So, a customer running the same version of the Business Suite on HANA will be running with a different database schema than a customer running on Oracle or DB2?

Jon said...
This comment has been removed by the author.
Jon said...

In my opinion they have to. Maybe not to start with (minimal disruption remember), but ultimately if Business Suite on HANA continues to use the same DB schema and application logic, then there will be little benefit.

Many important database tables in Business Suite are extracts / summaries of other tables which exist for performance reasons. These tables duplicate data and should not be required when running on a HANA platform.

In terms of the application logic, today most data processing is acutally done in the application server, not in the database server. This approach makes even less sense with HANA, so the application and database need to be rewritten to push data intensive operations to the database server.

My understanding is that this is exactly what is taking place with BW on HANA, with complex data operations (e.g. planning engine functions) being shifted from application server to database server, and over time I expect that the Business Suite will follow the same path.

clive boulton said...

Sage Group CSA Stephen Smith wrote about NoSQL ERP.

My take. Rather than SAP target rip-and- replace, innovating a NoSQL ERP hybrid with HANA can do more to disrupt Oracle (and probably Infor).

Matthew King said...

I agree that the schema needs to change - to avoid application integration headaches that all too often outweigh functional benefit and largely explains why so many customers are staying within the safe confines of the ERP core.

As to whether data processing needs to shift into the database I'm unsure. Application layer processing with hana is powerful because it too is in memory. But if true - then this deepens the bond between database and application - much like the bond between hardware and database in a so called database appliance. Good for SAP...

The OLTP-OLAP benefits are significant, but i think the operational benefits of "designing out" inferace complexities will be even greater. It's just that this is harder to do (read versus read-write architecture).

Ripping and replacing a database to save a few dollars is not reason enough to change. SAP needs to convince customers of the functional benefits...

Matthew King said...

This excerpt from vitalbi's post might explain the benefit of moving calculations into the database layer:

"Most of the data is stored in SAP HANA databases in columnar and compressed format. This data still has to be converted to records during processing, so it is important that this step happens as late as possible – something called late materialization. Ideally operations on the data should be able to run directly on compressed data, without need to uncompress them."