The Database Wars: Oracle vs Microsoft revisited

Till some time ago, software that manage

enterprise-wide data was known mostly by the convoluted acronym RDBMS for

relational database management system, or more generally DBMS. The RDBMS

business was relatively low on hype that surrounded other software like

operating systems and to take a more recent example-net browsers. This,

however, has changed rapidly in the past few years as businesses adapt to

the information age and learn to leverage that information to gain

competitive advantage. ERP, e-commerce and all other buzzwords one hears

today depend on data mining or data warehousing. In such an environment,

sophisticated database software takes centerstage. All of a sudden

'Oracle' and 'SQL' have become generic terms.

The market hotted up with the launch of

Microsoft's latest database product-SQL Server 7-late last year. SQL seems

to be the first database product from the Redmond giant that seems to

threaten Oracle Inc's supremacy in the database market. Close on the

heels, Oracle unveiled its 'database for internet computing'-Oracle 8i-in

February. Apparently, each company claims that its product has advantages

over the other. And, so for the SMEs or large enterprises planning to

implement or upgrade to a more capable database, it has become

increasingly difficult to cut through the hype and choose what best suits

their needs.

In fact, in the midst of the SQL 7 vs 8i

debate, it's easy to miss out on other options like IBM's DB2. "The

database market has seen its ups and lows. Vendors coming up and then

sinking without a trace. There was a time when Sybase and Informix were

big. But if you notice, in the background it's been Oracle and IBM

consistently. Oracle has been on the lines of Microsoft-playing more on

hype, big on publicity, while IBM has had a more traditional

approach," says Vanit Arora, Country Marketing Manager (IBM

Software), Tata IBM. IBM is the only player which is capable of

threatening both Microsoft and Oracle. Worldwide, according to Gartner

Group's Dataquest, IBM ended 1998 with 32.3 per cent market share, up from

28.9 per cent in 1997. While Oracle's share fell marginally from 29.4 per

cent to 29.3 per cent. The same report also showed Informix and Sybase

lose market share. The worldwide market which was estimated to be worth

$7.1 billion in 1998, is expected to hit $10 billion by the year 2003.

In India, however, Oracle is the leader by

far. In 1997-98, Oracle had a market share of 60 per cent by value of

shipments, followed by Sybase second with 18 per cent and Microsoft with

seven per cent. In 1998-99, Microsoft has improved its share to 19 per

cent and Oracle has lost share at 58 per cent. In conformity with the

worldwide trends, Sybase is losing market share in India as well. IBM's

DB2 in India has a small installed base in the government and is the DBMS

package of choice for private banks. However, DB2's government market

share is likely to fall in the wake of Microsoft tying up with the

National Informatics Center (NIC). The sale of DB2 is intrinsically tied

with the sale of the AS/400 systems and System 390 mainframes. The RDBMS

market in the country grew from Rs43.7 crore in 1997-98 to Rs59.0 crore in

1998-99 and is expected to be worth Rs75 crore by the year 2000-01,

growing at a CAGR of 34.5 per cent.

Technologically speaking



If the only consideration was sheer reliability and scalability, which
database product would one choose? Oracle, IBM, Informix and Sybase all

have been in the business much longer than Microsoft and from that stems a

technological lead. Features like row-level locking, multi-terabyte

capacity, replication and increased scalability, which Microsoft is

touting in SQL 7.0, have been there in its competitors' products for

years. The fact that SQL 7.0 has poor support for clustering and no large

table partitioning further shows that it has some way to go. However,

apart from the hype, perseverance characterizes Microsoft. SQL Server 7.0

is more of a contender than its previous version (6.5) ever was.

Oracle officials use the technology

argument for dismissing SQL 7.0 as not 'comparable' to Oracle 8. "SQL

7 is more equal to something like Oracle 6 (1988) or 7 (1991). On a

feature basis, it's impossible to compare the two (8i and SQL 7.0), they

are in two different leagues. Having said that, there is nothing that

Microsoft likes more than having the media, who don't really understand

these issues, explicitly comparing the two products. It's making their

product look better than it actually is," says Stephen Faris,

Director, Product Planning and Sales (South Asia Region), Oracle Corp.

Faris was in India to talk about 'state of the database' soon after

Microsoft launched SQL 7.0. However, Faris' views about the media

comparing the two databases seem to be inconsistent with the content on

his company's website (www.oracle.com).

The site not only has a comparison between the two products, but their is

also a quiz with prizes on offer and an image of a large, brutish Oracle

8i pulverizing a puny SQL 7.O in a boxing ring! "People are asking

the question, we are providing the answer," says Farris when asked

about the site.

Sanjiv Mathur, Marketing Manager, Microsoft

(India) Pvt Ltd, and point man for SQL Server in India, begs to differ

with Faris on the technology issue. "The Oracle engine is years old

and if you put a lot of things on top, it just makes it a layered product

and creates problems for people who have to manage all the new things in a

layered product." Adding another twist to the slugfest, Arora says,

"Microsoft SQL Server is essentially a bought out product. It was

Sybase, which they claim to have rewritten to a large degree. But if your

foundation itself is borrowed, how much can you rewrite on top of it? It

still has its limitations. Microsoft's style is typically consumer

friendly, but if you look at a mission-critical scenario in the

enterprise, they're still not there. The NT platform itself does not scale

up to that level."

As far as cross-platform availability is

concerned, the choice set shifts to all other DBMS products, as SQL server

runs only on NT. SQL server generally doesn't take advantage of more than

eight processors in an NT box while databases on Oracle, DB2 etc can scale

almost indefinitely on Unix. But this is a deliberate Microsoft strategy.

"The reason we do this is, first of all, a commitment to the Windows

platform, there's no denying that. The other is that this is the only way

that we can provide customers with a single product that always behaves

consistently. The third reason, the most important one, why we do it is

that if we were not to build this on a single platform, we would never be

able give customers a good consistent story of the tight integration that

we have between not only SQL server and the client applications but with

server applications, middle-tier applications, application development

tools and the office productivity applications," explains Mathur.

To TPC-C or TPC-D?



One way of comparing database software is by referring to the
benchmark tests conducted by Transaction Processing Performance Council (TPC)

of the US. The TPC has two main benchmarks hosted on its website,

www.tpc.org, one is an OLTP benchmark called TPC-C and second is a

decision support benchmark called TPC-D. The site describes the difference

between the two as: "Decision support is the capability of a system

to support the formulation of business decisions through complex queries

against a database. OLTP applications are update-intensive and generally

consist of shorter transactions that access a small portion of a database,

often through a primary key or index."

The TPC-C benchmark is further split into

results by performance and price/performance. The top ten TPC-C results by

performance have Oracle 8i occupying the first five positions based on

hardware supplied by Sun, Compaq, Sequent, HP and IBM, in that order.

However, when it comes to price/performance, all the top ten TPC-C results

go in favor of SQL Server 7.0 on hardware supplied by Dell, Compaq and

Unisys.

On the TPC-D benchmark, where results are

grouped by database sizes of 30GB, 100GB, 300GB, 1000GB and 3000GB, SQL

server does not figure. The fields are populated by the traditional

players-Oracle, IBM, Informix and NCR. But as with everything else, TPC-C

benchmarks, in particular, are controversial with each player interpreting

them in different ways. "On NT we are holding an absolute performance

record on TPC-C and TPC-D. Microsoft is not competing in those areas, it's

just running on the smaller systems. Now Microsoft would say that those

records are not relevant because they are on high-end systems. It's only

because they can't compete, that they are saying this," says Farris.

To this, Mathur reacts, "What Oracle tries to project with its

benchmark is that it has the most scalable and high-performance solution.

But, if you look keenly at what was used to get those benchmarks, one gets

a much clearer picture.

Oracle's best TPC-C benchmark on Windows NT

on a single server is on a 12 CPU machine. It loses squarely to SQL Server

7.0 on a single CPU Intel box on the TPC-C benchmark with its best

benchmark on a four-CPU machine. We have the better benchmark on single,

two-and four-CPU machines than any database on anybody's hardware using

anybody's operating system. Get me a four CPU Alpha box running digital

Unix and Oracle 8i, and we will lick them on the TPC-C benchmark. Their

next best benchmark on NT is on a cluster of six Compaq Proliant 6500

boxes with four CPUs each. If a customer wants a 12 CPU machine, I don't

have a better benchmark than Oracle. But if Oracle is clutching at straws

in the top 5 per cent of the market, so be it."

Farris says, pushing the entire big versus

small system debate under the carpet, "On the smaller scale systems

that are normally supporting smaller number of users, any of these

databases can more than adequately meet the peak performance demand. On

the higher systems, the absolute performance numbers are relevant, because

you typically buy larger systems for larger number of users." On the

issue whether SQL runs better on smaller systems, Farris says, "The

primary reason for the difference is that SQL is providing much less

functionality, so it has lesser overheads. The question is not who runs

better on a $1,000 computer. The key question is who provides the highest

reliability and availability which users are demanding. Who is supporting

the internet model of computing at the same time as client server, which

is where the market is today. Who is providing the ability to use the

latest technologies like Java? Where is Microsoft's support for

Java?"

Aah Java!



The technology seems to pop up everywhere. As far as Mathur is
concerned, Java support in SQL is not an issue, "Since we have one

platform that we run on, it doesn't make sense from a development

perspective to take a certain functionality and put it into individual

products. When you look at Microsoft's server-side applications, we are

talking about approximately 17 server products, only one of which is a

database product. When you look at Oracle's approach, it has three

server-side products-an OLAP server, a web server and a database server-so

they have to put it in the database."

The strategies



What threatens Oracle and the other RDBMS vendors is the tried and
successful Microsoft strategy of bundling its products or offering extras

at a price that is much lower than what the traditional players have been

charging. A case in point is the desktop applications market, which was in

the pre-Windows era dominated by players like WordPerfect (word

processors), Borland (DBMS) and Lotus (spreadsheets). As more people

adopted the Windows platform, Microsoft was able to leverage that

advantage to blast its competitors out of the water with its Office suite.

It has to be mentioned though, that Lotus, Borland and WordPerfect were

late to release Windows versions of their products for fear of helping

Microsoft establish Windows. Another similar example and the subject of

the US Department of Justice case against Microsoft, is the tactics that

the company used to gain share in the browser market.

With SQL, Microsoft bundles an OLAP

(on-line analytical processing) tool for free. SQL Server as a part of the

Back Office suite makes an attractive package for potential customers.

What Oracle, or for that matter any other database-only vendor, lacks is

the ability to offer a wide range of products that can be integrated on a

single platform like Microsoft and IBM can. With IBM, this advantage

extends to the hardware as well. The bundling works for Microsoft,

especially where large deals are involved or in the government sector they

can afford to undercut heavily.

Says Farris, "Microsoft's strategies

today are still designed to use combinations of low-price points and good

marketing messages to bring people in and then putting hooks into them and

before they know it they are locked into using their technology. For

Microsoft's strategy to succeed, customers and competitors must lose and

that's fundamentally different from our approach." He adds, "On

the one hand they accuse Oracle of charging different prices for the basic

and enterprise edition. Look at them, for NT they have three versions. In

fact, it's Microsoft whose pricing is predatory. They use pricing from

monopoly markets which they control-end-user OS and office segments-to

subsidize the pricing that they can offer in the enterprise market."

Claiming that SQL is cheaper to implement and upgrade on a per user basis,

Mathur adds that in terms of functionality, Oracle 8 Workgroup Server (the

standard version) does not allow the flexibility of adding on any extra

modules. "What you got is what you have, while SQL Server 7.0

standard edition includes the whole load of what we have."

With 8i and its Internet model of

computing, Oracle stretches the traditional client server model to a

three-tier architecture, which consists of a database server, application

server and a thin client. The model uses the internet/intranet as the

medium for distributing data throughout the enterprise. Oracle claims that

this model significantly cuts down on ownership costs by centralizing

application administration. However, Microsoft employees dismiss 8i as a

sales gimmick. Says a marketing manager at One Microsoft Way,

"Congratulations to the guy who put the 'i' behind the eight."

Adds Mathur, "Oracle 8i is an attempt to address a problem. It's

trying to address internet applications. Informix tried with universal

databases. IBM tried that with UDB as well. Oracle is trying that with

Oracle 8i. Essentially, they are the same thing, by putting a file system

inside the database they are able to store that information. But what am I

going to run that database on? On a file system or an operating system

that has a file system anyway? Can I run my web server off this file

system? It doesn't make sense."

But, the guy who gave us the infamous

network computer (NC) is at it again. One step beyond 8i is another one of

CEO Larry Ellison's ideas dubbed Raw Iron, where he has partnered with

fellow Bill basher Scott McNealy, CEO of Sun Microsystems. Raw Iron

envisages a dedicated PC Server running 8i on top of a Solaris microkernel

on hardware supplied by Hewlett-Packard. The idea behind Raw Iron stems

from the fact that half of all Oracle users run only the database on their

servers. By removing an OS like Windows NT from the equation, the

database, Oracle claims, runs much faster. The idea, of course, doesn't

cut ice with Mathur or Arora. One factor that could swing sales in the

long run is the success of the Linux OS. Linux is fast gaining marketshare

in the corporate space and is a popular choice for web servers.

Importantly, it is the only OS that threatens NT. Oracle, Informix and IBM

have launched or have plans to introduce Linux versions of their database

products. In fact, some international press reports have said that

Microsoft is toying with the idea of porting Office 2000 on to Linux and

perhaps even SQL Server. But denying any such reports Mathur says,

"All this Linux stuff is a gimmick. Linux is an open source software.

Anybody who imbibes this open source software culture must also make his

own code open source. Oracle is going to jump off this bandwagon the day

its customers start demanding free code from Oracle. It's nonsense, they

are not going to stand by that." So where does the entire debate

leave an IS manger looking to implement or migrate to a more capable

database? SQL 7 is the cheaper option and a very compelling one for sites

already running NT. Having said that, Oracle is the better bet for large

enterprise customers, mission-critical situations and non-Windows sites.

With its new mantra - the Internet changes

everything - Oracle may just be on the right track, but with Microsoft in

dogged pursuit, there is no room for error.