Object-Relational
Data Base Management Systems
ORDBMS:
Extending the Relational Model
ORDBMS:
Standards-Development Process
Real-world
ORDBMS Applications
Limitations
of the ORDBMS
Key
Players in ORDBMS
Object-Oriented
Data Base Management Systems
OODBMS:
A Single-level Store
The
OODBMS Standards-Development Process
Real-world
OODBMS Applications
Limitations
of the OODBMS
The
Future of the OODBMS
Key
Players in OODBMS
The Future
Changes in Technology
Convergence of Media
Legal Issues
Object-Relational
Data Base Management Systems
ORDBMS:
Extending the Relational Model
The relational data base model currently dominates the data base market
with approximately four billion dollars in sales per year.(1)
Changing user needs, however, dictate that object functionality will likely
be incorporated into most relational data base management systems (RDBMS)
in the near future. For example, consider a typical insurance company with
a customer data base and a claims data base implemented in RDBMS. The company
now wants this application to handle new data types such as diagrams of
accident sites, scanned images of police reports, pictures of damaged automobiles,
the location (latitude, longitude) of accident sites, the location (latitude,
longitude) of customers' homes, etc. The insurance company perceives that
by being able to query this new data, they will be able to make better
decisions, e.g., they will be able to identify the ten most dangerous intersections
and charge customers that live closest to those areas a risk premium. The
relational model, however, is not capable of supporting these new decision-making
capabilities. An object-relational data base management system (ORDBMS)
must be implemented.
ORDBMS is an extension of the relational model which allows richer data types to be supported. These new data types include user-defined abstract data types (ADTs), including image, audio, and video. Constructed types such as sets, tuples, arrays, and sequences are another example. A key feature of ORDBMS is inheritance which takes advantage of the commonality between different data types.(2) For example, voice data and music clips are different data types which have things in common. It is desirable to inherit properties of stereo music clips while defining voice data which may be recorded in mono.
For a more in-depth discussion of inheritance, click here: What
is Inheritance?
ORDBMS: Standards-Development Process
SQL92 is a standard which was created for relational data bases. SQL3
is now under development to support ORDBMS. It exists in draft form, and
ratification is anticipated in July 1998. SQL3 accomodates object SQL,
stored procedures, triggers, user-defined types, CLI (callable Level Interface),
senstitive cursors, and multimedia. The key point is that SQL3 handles
searching on complex data.
Real-world ORDBMS Applications
Oracle has recently introduced Oracle8, an ORDBMS designed to "manage
large amounts of information securely, reliably and economically over computer
networks. It is a versatile information platform that enables high speed
transactions, better business decisions and sophisticated object relational
applications."(3) Some interesting
examples of Oracle8 data bases include:
Abbey National, the United Kingdom's fifth-largest
bank, is developing a new customer data base system to improve manageability
of large objects, such as Abbey National's Financial Transactions data.
Abbey anticipates using Oracle8 objects to store more complex forms of
data, such as other data bases and tables. An individual's complete insurance
record, including insurance policies, claim forms, credit card receipts,
and photos of automobile collision damage, can be stored as an object,
providing a view of data that approximates the traditional paper record.
Nasdaq Stock Market has been testing Oracle8 on
Sun Solaris and Sequent Dynix platforms to support its large Decision Support
System (DSS) data bases. Oracle8 is expected to provide Nasdaq with robust
performance, advanced replication, and segment partitioning to help Nasdaq
manage its Very Large Data Bases (VLDBs), with Oracle8 scaling up to handle
multiple terabytes and 30,000 concurrent users, providing substantial headroom
for Nasdaq's hundreds of gigabytes.
Esther Dyson, a leading technology futurist, describes a primary limitation of ORDBMS as follows: "Using tables to store objects is like driving your car home and then disassembling it to put it in the garage. It can be assembled again in the morning, but one eventually asks whether this is the most efficient way to park a car."
When developing a Web application using a relational data base, the
data models in the data base and the application are incompatible. This
requires a conversion of data from relational to object every time the
application accesses data. It is this constant conversion of data from
relational to object and back again which results in poor performance.
The underlying architecture of the OR model renders it wholly inappropriate
for developing high-speed Web applications. With an object data base, the
same data model exists in both the data base and the application. Here,
the data can be passed between application and the data base engine without
the overhead inherent in the object-relational model.
The "Giant Vendors" of RDBMS expanding into ORDBMS: (4)
IBM (DB2 version 3): http://www.software.ibm.com/data/db2
Informix (acquired Illustra):
http://www.informix.com/infmx-cgi/Webdriver?MIval=dynamic_server
Microsoft: http://www.microsoft.com
Oracle (Oracle8): http://www.oracle.com/corporate/pressroom/html/o8factf.html
Sybase (Adaptive Server - introduced (9/97): http://www.sybase.com/adaptiveserver/
Other ORDBMS Vendors:
Cloudscape's JBMS (just released--a java-based ORDBMS):
http://www.cloudscape.com/fs-prod.html
Hewlett-Packard: http://www.hp.com/sesd/3rdparty.1/object.database.html
UniSQL: http://www.unisql.com
Unisys: OSMOS: http://www.osmos.com
Object-Oriented
Data Base Management Systems
The Object-oriented Data Base Management System (OODBMS) is based on a single-level store of objects, facilitated by CORBA's Persistent Object Service (POS). POS provides a single client interface for storing objects, regardless of how or where the object is stored, including file systems, relational data bases or object data bases. In a single-level store, there is no differentiation between memory and persistent storage. It is for this reason that the OODBMS provides, in contrast to the ORDBMS, the most direct path for storing objects, lending the best performance and versatility to the object model data base. With the OODBMS there is no impedance mismatch. (The ORDBMS, on the other hand, contains a very big impedance mismatch. The exchange path of the ORDBMS is not simple, which accounts for the lack of efficient performance in the ORDBMS.)
The OODBMS is based on a client-server architecture: data is shared through the management of persistent objects. OODBMS use a superset of the OMG Interface Definition Language (IDL) to describe their objects--objects are thus represented in a very clear manner.
The OODBMS allows optimal control of complex data and complex interrelationships among objects. Some real-world applications that are best suited to the OODBMS include financial portfolio risk-analysis systems; telecommunications service applications; WWW document structures; design and manufacturing systems; and hospital patient record systems.
An advantage to the OODBMS is that data base capabilities are integrated
directly into the object-oriented programming language (C++, Smalltalk,
Java). This presents a different structure than the relational or object-relational
model that uses a separate language (Structured Query Language, SQL) to
define, retrieve, and manipulate data. Drawing on the client/server architectural
model, with the OODBMS, the client is, for example, the C++ program and
the server is the OODBMS.
The OODBMS Standards-Development Process
The Object Data Base Management Group was established in 1991 to develop the ODMG standard. The goal of the ODMG was not to define a completely new standard from the ground up, but to build on existing standards, including OMG, SQL-92, and ANSI programming language standards, to define an application-portability framework among OODBMS. The ODMG standard was established as the answer to SQL--ODMG permits the OODBMS to query and update data base objects.
The three major components of the ODMG-93 are:
Object Definition Language: ODMG-93 uses
the OMG IDL as its data definition language. ODL is a superset of IDL because
it defines elements that are not in IDL, such as collection classes and
referential relationships. ODL provides interface and data definition portability
across languages and OODBMS vendor platforms.
Object Query Language: ODMG-93 makes use
of a SQL-like declarative language for querying and updating data base
objects. It supports the most commonly used SQL functions, including joins,
but purposely leaves out most of SQL3 semantics because of perceived limitations
in its data model. To expand broad usability, OQL and SQL3 may converge
in the future.
C++ and Smalltalk language bindings: ODMG-93
defines how to write portable C++ or Smalltalk code that manipulates persistent
objects. The developers of ODMG-93 did not believe exclusively in a "universal"
Data Manipulation Language, like SQL. Instead, they proposed a "unified
object model for sharing data across programming languages, as well as
a common query language." They wanted to respect the syntax of the base
language into which it is being inserted. According to an engineer at Object
Design, "This enables programmers to feel they are writing in a single
integrated programming language that supports persistence."(5)
In July 1997, ODMG 2.0 was published to offer three major improvements: better C++ and Smalltalk language bindings; a new ODMB binding for Java that standardizes and simplifies the storage of Java objects in data bases; and a metamodel that provides a language-independent description of a data base schema to tools and applications.
The ODMG membership consists of voting and reviewer members. The voting member companies are GemStone Systems, IBEX Computing SA, O2 Technology, Object Design, Objectivity, POET Software, UniSQL and Versant Object Technology. Reviewer member companies are Andersen Consulting, CERN, Electronic Data Systems (EDS), Fujitsu Software Corporation, Hitachi, Lockheed Martin, Microsoft, MITRE Corporation, NEC Corporation, ONTOS, Persistence Software, Sybase, Unidata and VMARK Software.
Voting membership requires that a company must have developed and must commercially ship a object DBMS product. The company must commit 20 percent of the time of a senior object DBMS expert and must commit to implement the ODMG standard. Reviewer membership requires a company to commit 10 percent of the time of a senior object DBMS expert, but does not require that the company ship an object DBMS product. (Participation from major companies like Microsoft and Sybase as reviewer members indicates the potential mainstream development and use of the OODBMS. It is interesting to note that at this point in time, however, Oracle remains committed to the relational and object-relational model.) The ODMG has also recently created a new certification membership, which allows corporate end users, consultants, trainers, and tool vendors to participate at a lower level of commitment.
Real-world
OODBMS Applications
ObjectStore is an OODBMS application developed by the ODMB voting member
Object Design, based in Burlington, Massachusetts. ObjectStore is a popular
OODBMS application, and was voted DBMS Magazine's 1997 Reader's Choice
for Best Object-Oriented Data Base.(6)
ObjectStore is used in several interesting applications, which help to
illustrate the OODBMS in action. Some interesting examples of ObjectStore
data bases include:
The Human Genome Project at the MIT Whitehead Institute
(Cambridge, MA) is helping to sequence the entire human genome by the year
2005. They have two applications: MAPBASE (genetic mapping of mice) and
LABBASE (physical mapping of humans and mice), both of which use ObjectStore
as the data and model repository. They chose an OODBMS for the repository
because of its ability to efficiently and quickly manage complex genetic
data.
Monterey Bay Aquarium Research Institute uses ObjectStore
to collect and store multi-media oceanographic research data (videos, image,
etc.). They share their ObjectStore-based research data with other research
organizations through an extranet application written in Java. They replaced
Oracle with ObjectStore because ObjectStore reduced the amount of Java
code required (no need to convert their objects to rows & columns)
and because ObjectStore could be extended very easily to store any type
of multi-media or textual data.
Rush-Presbyterian-St.Luke's Medical Center is a
1,000-bed research hospital in Chicago. ObjectStore supports a patient
record system for their echocardiography unit. Doctors study echocardiograms
(ultrasound recordings of hearts beating) and enter their patient evaluations
into ObjectStore. The patient's report is automatically generated in HTML
and FAXed to the patient's hospital room in 15 minutes (hours faster than
the old, paper-based system).
Versant Object Technology, another voting member of the ODMG offers
another interesting example of an OODBMS:
NUSTAR International (Atlanta) uses Versant to accept,
store, manage, and publish data about real estate in many markets across
the country. OODBMS allows great flexibility for updating local editions--each
of which includes photos and other information about homes for sale in
a given area. Publications are updated very frequently. Company scans between
16,000 and 25,000 photos a week. Each photo is associated with a real estate
listing that includes text and graphics (including maps, floor plans, or
logos).
Some of the commonly recognized limitations of the OODBMS include: OODBMS
are limited to small applications; OODBMS don't support a lot of concurrent
users; and it takes too much time to deliver an OODBMS application. To
investigate the validity of these charges, the ODMG conducted a study this
year of 24 OODBMS applications. They found that although these limitations
exist, the capabilities of the OODBMS are expanding. For example, regarding
the view that OODBMS only handle small applications, of the 24 applications
examined, 15 occupied between 1 and 99 GB, a pretty sizable data base.
In regards to concerns about concurrency limitations, 4 applications were
found to support more than 500 users, with the most common situation involving
between 10 and 99 users. (Compared to the NASDAQ ORDBMS developed by Oracle8,
however, which supports 30,000 concurrent users, the OODBMS is not the
appropriate model for data bases that require large concurrency.) And in
regards to development time, it was recognized that development can be
time-consuming, but not in every case. Delivery time is improving, which
is facilitated by the greater availability of vendor tools. Of the applications
studied, the average development period was between twelve to twenty-three
months, but many companies delivered in less than a year.
The market for OODBMS is growing fast, but is still dwarfed by the market
for relational and object-relational data bases. In 1995, the market for
OODBMS drew $100 million. It is predicted to grow to $430 million by 1997
and $600 million by 2000.(7) (In contrast, it
is predicted that the market for ORDBMS will grow to $1 billion by 2000).
The test will be time. In ten years, if object-oriented programming becomes
the most commonly used model of programming, and if the identified limitations
of the OODBMS are overcome, then we can probably anticipate that the OODBMS
will be more widely used.
Key Players in OODBMS: Standards-Making Organizations and Vendors
The Object Data Base Management Group (ODMG): http://www.odmg.org
Voting Members of ODMG:
GemStone Systems Inc. (Beaverton, OR): http://www.gemstone.com
O2 Technology Inc. (Palo Alto, CA): http://www.o2tech.com
Object Design (Burlington, MA): http://www.odi.com
Objectivity Inc. (Mountain View, CA): http://www.objectivity.com
(See especially their explanation of object data
bases
at http://www.objectivity.com/ObjectDatabase/why_odb.html
)
POET Software Corp. (San Mateo, CA): http://www.poet.com
Versant Object Technology Corp (Menlo Park, CA): http://www.versant.com
Other OODBMS Vendors:
Cincom Systems Inc. (Cincinnati, OH): http://www.cincom.com
Computer Associates International Inc. (Islandia, NY): http://www.cai.com
Informix Software Inc. (Menlo Park): http://www.informix.com
Mainstay (Camarillo, CA): http://www.mstay.com
NeoLogic Systems (Berkeley, CA): http://www.neologic.com
Ontos Inc. (Lowell, MA): http://www.ontos.com
Raima Corp. (Issaquah, WA): http://www.raima.com
Rational Software Corp. (Santa Clara, CA): http://www.rational.com
Secant Technologies Inc. (Beachwood, OH): http://www.secant.com
Speculating about the future of technology is usually futile. Predictions
have been notoriously inaccurate. The future of data bases will be influenced
from three directions: changes in technology, the convergence of media,
and legal issues.
Traditional relational data bases will remain the DNA of business for
commerce and analysis require speed of access and large storage capabilities.
However, digital convergence and Web communication will create an emergent
layer of commerce and culture that can only be stored,
manipulated, and analyzed by object data bases. Hence, the object-relational
data base will become the heart and pulse of the database industry. Relational
data bases have features that make them the natural choice for storing
the continuing accumulation of large amounts of data. Object data
bases have features that offer users malleability and customization. The
marriage of the two results in the large storage capacity and access speed
of the relational data base combined with the manipulation options of the
object data bases. These features, combined with what Fortune Magazine
calls "a radical shift occurring in corporate computing -- think of it
as the recentralization of management"(8), make the object-relational
data base the most likely winner in this evolutionary struggle.
The object-relational data base is the best bet of the data bases extant
today. But there are two trends that will stretch the abilities of current
data bases and probably lead to data base innovation. One is the convergence
of television and computers; the other is the increased use of audio.
These two trends are reflected by Bill Gates recent investments - $1 billion
in the Comcast TV-cable company and $425 million to buy WebTV Networks.
At the same time he is investing in research because, he says, "the future
of computing is the computer that talks, listens, sees, and learns."(9)
While Mr. Gates is not a trend-setter in creating new technologies, he
certainly popularizes and markets them. The computer and television
convergence is already here. "Today, "Net-top" devices from
WebTV, Curtis Mathis, and RCA, as well as hybrid computer-televisions,
such as Gateway 2000's Destination and Compaq's PC Theater, let viewers
surf the Web while watching TV."(10) Companies
such as Speech Systems, Inc. are developing voice recognition programs
that use databases to store not only grammar and syntax but also to learn
new formulations from the user. "The software handles sentences …on
the fly, the computer displays its best guess as the dictating continues
until a final form is reached. Then, if the computer guesses wrong, the
software accepts corrections from the user…and adds that particular sentence
to the database for future reference."(11)
The on-going discussion between the European Union and the United States
over the protection of data base information may also play a role in the
future evolution of data bases. "At first glance, the material reciprocity
clause of the European database directive might seem to threaten the interests
of U.S. database developers by denying them the benefits of the European
sui generis right in database contents if the U.S. does not adopt an equivalent
law by January 1998."(12) The European Union
is proposing very strong, continuing copyright protections for data base
information and requiring reciprocity agreements from other nations. The
United States does not recognize "sweat of brow" copyright on data bases
and is not likely to pass a similar agreement. There has been some speculation
that "European companies are lying in wait for January 1998 in hopes of
sucking all of the valuable data out of U.S. databases"(13)
Professor Samuelson considers this possibility patently absurd and offers
numerous points in support of her argument. However this particular
piece of legislation is resolved, legal issues and international law may
well play an important role in the future development of data bases.
Endnotes
1Michael Stonebraker. "Object-Relational DBMS--The Next Wave," Informix website.
2Hellerstein, Joseph M., Chapter 21: "Object-Database Systems," in Ramakrishnan, Raghu: Database Management Systems (McGraw Hill, 1997).
3 http://www.oracle.com/corporate/pressroom/html/o8factf.html
4Kim, Won.
"Bringing Object/Relational Down to Earth," Database Programming &
Design, July 1997, pp. 26-35.
5Tom Atwood quoted in Orfali, Robert,
Dan Harkey, and Jeri Edwards. The Essential Distributed Objects Survival
Guide, (John Wiley & Sons, Inc.: New York, 1996), p.161.
7Statistics from the Cowen Research Report, Sept. 15, 1993, and the Data Analysis Group, Feb. 7, 1997.
8Martin, Michael H. and Stipp, David. "10 Tech Trends to Bet On," Fortune Magazine. 11/10/97.
9Moore, Alicia Hills. "Mr. Gates Builds His Brain Trust," Fortune Magazine. 12/8/97.
10Thomas, Susan Gregory and Egan, Jack. "The Networked Family," U.S. News and World Report. 12/1/97.
11Coates, James. "It Takes Trial and Error to Teach a PC to Listen," Chicago Tribune. 12/1/97.
12Samuelson, Pamela, in a letter to Representative
Howard Coble, Part B. 10/22/97.
13Ibid
Sources
Barry, Douglas K. "The Truth About Object Databases," Database Programming & Design, July 1997, pp. 44-48.
Coates, James. "It Takes Trial and Error to Teach a PC to Listen," Chicago Tribune. 12/1/97.
Communications of the ACM, Special Section: Next Generation Database Systems, October 1991.
Davis, Judy. "Extended Relational DBMSs: The Technology, Part 1," DBMS Online, June 1997.
Frank, Maurice. "Debating Databases at DB/Expo," DBMS Online, December 1996.
Hellerstein, Joseph M., Chapter 21: "Object-Database Systems," in Ramakrishnan, Raghu: Database Management Systems (McGraw Hill, 1997).
Jakobovits, Rex. "Comparison to Relational Databases."
King, Nelson H. "Object DBMSs: Now or Never," DBMS Online, July 1997.
----. "The Object Database Goes Online," Internet Systems, January 1997.
Linthicum, David S. "Selecting a DBMS," DBMS Online, July 1996, http://www.dbmsmag.com/0607d11.html
Martin, Michael H. and Stipp, David. "10 Tech Trends to Bet On," Fortune Magazine. 11/10/97.
Messerschmitt, David G. Chapter 11: Middleware, in Distributed Computing from an Application Perspective (forthcoming).
Moore, Alicia Hills. "Mr. Gates Builds His Brain Trust," Fortune Magazine. 12/8/97.
Orfali, Robert, Dan Harkey, and Jeri Edwards. The Essential Distributed Objects Survival Guide, (John Wiley & Sons, Inc.: New York, 1996), pp. 151-164.
Samuelson, Pamela, in a letter to Representative Howard Coble, Part
B. 10/22/97.
Cook, Rick. "Is
a hybrid database in your future?" SunWorld: Object-relational databases,"
February 1997.
Stonebraker, Michael. "Object-Relational DBMS--The Next Wave," Informix website.
Thomas, Susan Gregory and Egan, Jack. "The Networked Family," U.S. News and World Report. 12/1/97.