META Trend:By 2005/06, Intel-based servers will dominate at the DBMS tier. Scale-out database configurations will grow in popularity as management virtualization improves, but remain primarily a high-availability option. By 2004, XML-based content storage and rendering will become a key DBMS differentiator along with data movement and transformation. Centralized data warehouse architectures will emphasize database workload prioritization and data archiving to support the increasing need for near-real-time analytics through 2006.
Although much has been written about the price disparities between Oracle and other DBMS offerings, our research points not to high prices (especially since the pricing change in June 2001), but rather to general confusion over Oracle's licensing policy as the source of most customer frustration. Through 2007, the new landscape of Web-based applications and application integration via Web services will render user-based licenses extinct.
By 2003, we expect Oracle to either drop user-based pricing as an option or make moves to remove the ambiguities around user-based licensing. Value-based pricing, in general, will not withstand user scrutiny; Oracle must take steps within the next 6-12 months to remove licensing as an issue for users, or risk an irreversible slide in market share as organizations adopt lower-cost Intel-based servers running Windows, Linux, or other Unix variants.
Ultimately, we believe database licensing will evolve into a pure capacity-based (i.e., processor) model for all vendors as a means of eliminating "gaming" by some users and simplifying terms and conditions for everyone concerned. The only other alternative would be some type of utility-based pricing, in which the actual usage is metered and a monthly bill is calculated by the database software. However, issues exist on both sides concerning this method of licensing. Vendors are concerned with adapting existing commission structures to a utility pricing model, while users are concerned with possible fluctuations in usage, making budgeting difficult. The best way to approach such an issue is for vendors to put this capability in place and enable users to view what potential costs would be.
Q. How is the term "multiplexing" typically defined in a database context?
A.In a database context, the term refers to the sharing of a pool of database sessions by many users. It is commonly used in n-tier applications, where an individual user connects via a middle tier (e.g., via a Web server or TP monitor) to an application. The application connects to the database on behalf of the application, which may maintain a pool of connections to the database shared by all users of the application. From the database perspective, only a small number of users (application transactions/sessions) are logged in, either concurrently or by having been granted the privilege of connecting to the database.
Q. What is different about how Oracle is interpreting the term "named user," or multiplexing?
A.Previously, Oracle defined a named user as "an individual authorized by you to use the program"; this included the fact that any "non-human operated device will be counted as a Name User." Generally accepted practice is that batch loads from one application database into another are considered a "non-human operated device," because the connection was through a program with no user intervention and, therefore, licensed as a single named user.
According to Oracle, the previous language on multiplexing and/or non-human use of the database needed "clarification," so a document provided to META Group clients titled "Topic: Database Licensing" described several licensing scenarios and stated that "any automated process that involves a computer-to-computer connection must be licensed using the processor metric. Oracle treats any batch process as a multiplexor." In discussions with META Group, Oracle maintains that there has been no change to its licensing; rather, it has simply been a clarification.
Q. What is the impact on Oracle customers as a result of this "clarification"?
A. We estimate that 70% of our clients owning Oracle licenses own some named-user licenses or older concurrent user licenses. Every such client is at risk for Oracle to claim that the company is potentially out of compliance. On this basis, companies could be pushed by Oracle to convert to the capacity-based licensing model. This hits Oracle data warehouse (DW) customers particularly hard, because virtually every DW receives batch feeds from other systems as part of regular ETL (extract, transform, and load) processing. The very nature of a DW (large databases but few users) has made named-user pricing attractive and, because Oracle's main competitors (e.g., IBM, Microsoft) offer predominantly capacity-based licenses, Oracle has capitalized on this differentiation. Oracle's recent TPC-H (data warehousing) benchmark plainly illustrates this advantage (see Figure 2).
Many IT organizations support legacy applications based on technology other than Oracle's database (e.g., IDMS, IMS, DB2 UDB OS/390, AS/400). Under Oracle's clarified multiplexing language, even having a single non-Oracle source system may require an organization to license the database using CPU pricing, or to pay for all the users of the source system. This essentially ends named-user pricing as an option for many of Oracle's customers and prospects. As illustrated by Oracle's TPC-H benchmark, the cost for licenses alone could increase as much as 500%, even after discounts are applied.
Q. How can Oracle justify such an argument?
A.Oracle believes a company should pay for "the value" the database delivers to the business. For example, a utility company automatically collects meter data from 500,000 customers and stores it in a proprietary database for billing purposes. Once a week, data is summarized and loaded into an Oracle database for analytic reports that are created by 20 different users. Oracle holds that customers of the source application (i.e., every utility customer with a meter) receive value from the Oracle database and should be licensed - in addition to the 20 named users who have the access rights to query the Oracle database.
We believe this argument is without reason. If we are to accept that the utility customers of our example gain value from Oracle (and, we assume, should see a line-item charge on their bill for the $800 named-user license), then we must also include all 300 million people in the US if the company receives a tape from the US Census Bureau. The database is a tool and, as such, only provides value to those who use the tool directly.
Q. What should a company faced with this situation do?
A. First, if approached, users should escalate to the highest levels of management and to individuals dedicated to the task of asset management. Second, users should conduct an informed contract review and a thorough audit of Oracle usage. Third, users should speak with the Oracle officer that is responsible for licensing and pricing to obtain clarification; they should not rely on the Oracle sales rep. Fourth, users should firmly reject Oracle's claims and demand that Oracle state any claims of non-compliance in writing.
If forced, organizations should take legal action to protect their interests. In addition, we estimate the average Global 2000 organization to be 30% overlicensed, at a cost of $200K-$600K annually. This occurs via methods such as frontloading user licenses, underutilizing CPUs (if using capacity-based pricing), or simply licensing the wrong version of the database. Research indicates that 60% of databases run on servers of four or fewer processors. This means that many could license Oracle's standard edition ($15K per CPU, or $300 per named user) and save substantial sums on licensing and maintenance.
Business Impact: License maintenance is part of an overall infrastructure portfolio management process. Standardized and measurable licensing models result in better management and lower costs.
Bottom Line: For all practical purposes, Oracle's "clarification" of multiplexing to include batch loads essentially ends the named-user pricing model. Organizations being pressured by Oracle to convert to capacity pricing based on this change should seek legal counsel.
META Group of Stamford, Conn., is a leading research and consulting firm, focusing on information technology and business transformation strategies. For more information, visit MetaGroup.com.