Columns
Q&A
N>obody knows data like Janet Perna. Over the last few years, Perna, general
manager, data management, IBM software solutions, has been IBM's point person
in this crucial arena. ADT Editor Mike Bucken and Managing Editor Jack Vaughan
talked with Perna recently about XML, intelligent mining, universal databases
and a few other items that are shaking up this slightly mystifying technology
segment.
How do you rate the status of the Universal Database?
If you look at what we've done with the Universal Database, we've extended
relational technology to support other data types, other kinds of functions,
and to provide an extensible relational database. That's one of the things we've
done with it. We've [also] provided support for data that's text or image or
audio or video, or you could think up things like XML. That's one aspect of
the Universal Database.
But there are other aspects of Universal Database that have to do with scalability
-- being able to scale from small systems to very large systems, including massively
parallel systems. Universal in terms of the types of applications -- supporting
those online transaction processing kind of applications -- as well as business
intelligence, which requires complex query and analytic capability. Things like
universal access -- access to the database from the Web, and then reliability
and manageability of the environment.
When we talk about the Universal Database, it's not something that's different
than relational, it's an extension of relational technology. A growing number
[of customers] are using it as a base for a business intelligence solution,
warehousing platform or data mart platform. And there are others who are using
it as a base for their e-business applications, like electronic commerce. When
you start thinking about a base for e-commerce, you're thinking about application
requirements that require very high reliability, availability, scalability and
performance, things that IBM databases are hallmarked for.
Do people still buy it strictly as a relational database?
I think most people have continued to use it to manage structured data if you
will --character type of data, alphanumerical data, that kind of data. There
are a growing number of customers who are using it for things like managing
text. If you look at Sapien Health Network, for example, they've got a Web site
that enables visitors to their Web site to do text searches and to find information
on different diseases. And they're using the Universal Database with a text-extender
capability to provide that.
Is IBM holding up its end in the meta data field?
Meta data is a topic that's been discussed in the industry for a number of
years. And getting a standard meta data format. From a customer point of view,
it's something that would clearly benefit customers as they try to integrate
data from various vendors and applications. In that vein, IBM has taken a leadership
position, along with a couple of other companies within the standards community,
to work on some meta data issues using XML as a common meta data exchange format.
We have a lot of work going on in the meta data area. I think that at least
most of the vendors we are working with see the potential of XML as a way to
do common data interchange.
Do you have a gauge on IBM's thinking on repositories?
As you probably know, IBM does have an application development repository in
Team Connection. And again, what our approach is there is really to extend that
and to be the meta data repository for data, as well as application development.
Lotus and IBM had quite different data strategies. How is the integration
going?
Well, I think we've achieved quite a bit with Lotus in terms of data integration.
We've opened up DB2 data to Notes applications, so if you're writing an application
using the Notes APIs, you can access DB2 data. You can integrate that data and
extend Notes applications with relational. We also have a capability to extract
data from DB2 and populate Notes databases with DB2 data. If you think about
it, it's a marriage of structured and unstructured data providing application
developers with the ability to integrate through the same API relational data
with the Notes data. And that work has been completed, by the way. We shipped
DB2 for Domino last year, that enables our capability. We're working with a
number of Notes application development vendors; we want to extend their Notes
applications with relational capability.
It seems like search technology is almost overshadowing database technology
these days.
There are a number of technologies that we're working on and Search is one
of them. We shipped a product called Intelligent Miner for Text that provides
advanced search, text-search and text-mining capability, as well as a Web-crawling
capability. We've taken the search technology from there and Lotus has incorporated
the search technology -- the advanced search technology -- into Domino. It's
an instance where we're applying technology that initially came out of IBM research
to now incorporate it into two different products -- one from Lotus and one,
the Intelligent Miner, from IBM. By the way, we have a lot of experience in
providing scalable, high-performing relational databases and we're applying
a lot of the technology and experience that we've gained through the years in
DB2. We're applying that to Notes to be able to scale the Notes environment.
Some RDB vendor growth is slowing. Does the world have enough relational
databases?
We've had a tremendous first half. Well, actually, the first three quarters
have been tremendous for us, but let me recap the first half of this year. IMS
and DB2 on the mainframe each grew 14% in the first half of the year, year-to-year.
That is quite a bit faster than our competitors. DB2 on the distributed platforms,
Unix, NT and LS2, grew 68% year-to-year in the first half. So we're seeing good
growth there now, whereas our competition is seeing a lot less growth.
What we're also seeing is -- and this gets back to the question about what
are people using relational databases for -- that a lot of this growth is coming
from business intelligence kinds of applications. Platforms for data warehousing,
data marts, complex decision support types of processing, all of these solutions
require relational databases. And they require a very strong query optimizer.
As you scale up on them, they require a parallelism in the query optimizer.
This part of the market for relational databases is projected to grow around
30% between now and the year 2002. And so [people are] doing more and more of
that type of application. And we're seeing a lot of growth in that area.
Why DB2?
If you look at the history, Oracle and Informix grew very, very quickly as
customers began to implement packaged solutions -- ERP-type applications on
Unix. The whole client/server movement provided Oracle, Informix and Sybase,
to some degree, with a big growth potential over the last seven years. IBM didn't
have a product on the Unix platform until the end of 1994. We missed a lot of
the growth that occurred in the packaged applications space for online transaction
processing during that period.
Since 1994, we've had competitive DB2 products on Unix platforms. We shipped
on NT in 1995, and we've had parallel query capability and very scalable parallel
query capability with DB2 cross-platform since 1995 as this market's taken off.
We're seeing quite a bit of growth coming from the uptake in that market and
the fact that we do have a very competitive product available on those platforms.
By the way, a lot of DB2 growth on the 390 is also coming from these types of
applications for business intelligence.
Why would the Informix Universal Server fail so badly?
I don't know. They had good parallel technology. Informix does have good parallel
technology. I don't know. I don't think it was necessarily a technology issue.
How much business are you seeing on non-IBM platforms?
We're seeing growth on the platforms. Am I satisfied? I'm probably never going
to be satisfied. As I said, we're seeing good growth on the Sun platforms. We're
seeing good growth there again for business intelligence, the same with the
RS/6000. We're providing customers with platform options to deploy their database
solutions.
In fact, in our developer's kit for DB2 we package VisualAge for Java; VisualAge
really is the application development portion of DB2. We supported Java, stored
procedures and JDBC very early on with DB2 -- almost two years ago. A key part
of our application development strategy, as you know, is Java support. And it's
very important that we have VisualAge Java support that works well with DB2
as the underlying database.
What is your position on SQL 3?
We're supporting elements of SQL 3 already, so if you look at what we've done
in the object relational area, a lot of those extensions are extensions to relational
that are in SQL 3. We've also written many of the papers that have gone to define
the SQL 3 standard. We have implemented parts of SQL 3 as early as Version 2
of DB2 -- which was our 1995 product -- and we continue to take elements of
SQL 3 and implement them.
Particular things, like the object relational support, are important as we
link together our application development strategy with our database strategy
because the object relational support in relational databases enables us to
do things like model Corba objects in relational. That gets us support in enterprise
JavaBeans. It's a close melding of our application development strategy with
database.
Is there any growing competition from object databases?
I still see them in the niche that they're in. In fact, they don't appear to
be growing as much as they were even five years ago.
What do you attribute this to?
I think the relational database players are all moving to extend relational
with objects. And you kind of look at how you get something to go mainstream,
and the applications aren't there for the object relational databases. The application
vendors are supporting the relational database spenders, and to the degree that
we can provide object support in relational databases, it extends their application
investment in relational. It enables them to use object-oriented programming
tools. I don't see that changing in the near term.
I think one of the key points about what people are using relational databases
or the DB2 Universal Database for, [is that] it's the same stuff. It's online
transaction processing. There's more growth in business intelligence now, which
is driving a lot of our growth for those new application types. And, of course,
the Web is going to drive additional growth, particularly as relational databases
begin to manage rich content. So text, video, audio, XML, that kind of data.
And again, what we're doing with the DB2 Universal Database is extending it
to provide that capability. One of the other things that we're doing, and that
we'll be shipping before the end of this year, is our Data Links technology
that enables you to link an external file to the relational database and provide
referential integrity.
How is database revenue stacking up?
IDC has published some revenue numbers as has Dataquest, which are roughly
accurate. If you look at the whole DB2 family of products it's over a billion
dollars. And if you look at market share according to Dataquest, we're neck
and neck with Oracle at a little over 27% market share. If you look at Sybase
and Informix, they're less than 5%.