—Ross
Templeton,
Program Manager (Datawarehousing and
High Availability Systems), Asia-Australia Region, Hewlett Packard.
Soft-spoken and suave, Ross
Templeton heads Datawarehousing and High Availability Systems division [HAS] at HP in the
Asia-Australia region. On a recent visit to India, concerning a seminar by HP and its
partners on the same subject, he met with DATAQUEST for an exclusive interview to discuss
datawarehousing and HAS, the analysis that is possible with datamining tools, the
implications for India, and other associated issues like platforms and operating systems
best suited for the task on hand. Excerpts:
What is HP’s strategy for
datawarehousing?
HP takes an open systems approach. We work with our partners to provide solutions, and
integrate them for our customers. We have several pre-certified solutions readily
available, which are actually partial solutions which require customization. Our partners
do the customization. We also provide consulting services to help our customers leverage
on datawarehousing solutions.
What sort of High Availability
Systems (HAS) do you offer?
We try to ensure that systems are available as much as possible. This we manage with
hardware components like RAID disks and the like. We offer clustering solutions and other
solutions based upon the MC/Service Card to ensure high availability. We also have
consultancy services which help our customers develop HAS architectures.
Datawarehousing concerns analysis,
while HAS concerns mission-critical applications. Yet, they appear to get clubbed by the
likes of HP and NCR. Why is this so? What link exists between them?
No, there is no direct link between HAS and datawarehousing, though very often,
particularly in large datawarehousing solutions, there is a need to ensure data safety.
Still, there is no reason to ensure high availability in most cases.
What sort of analysis are people
doing with datawarehousing? Can you mention any success stories of datawarehousing
solutions from HP?
The main industries which require datawarehousing are industries like Telecom, which use
datawarehousing for analyzing information on individual telephone calls to determine
high-value customers, unprofitable customers, turnover of customers to the competition,
flaw detection, and network performance monitoring. Financial institutions, mainly banks,
use datawarehousing to generate detailed marketing analysis to help them better understand
their customers. Other industries include the retail industry, where datawarehousing is
used by supermarket chains to determine what products to sell together, and to track how
products are moving on a daily, weekly, or monthly basis.
Is India ready for datawarehousing?
I think that telecom companies and some of the large manufacturing companies in India are
ready for datawarehousing solutions. Since the banks are mainly Government owned, there
may not be competitive pressures which encourage them to get into datawarehousing.
Talking about banks, the Narasimham
Committee on banking sector
reforms has recently suggested steps to make some banks more internationally
competitive…
International competition is one of the drivers of datawarehousing. Other factors which
make it required include deregulation, globalization, and new technologies like the
Internet. Deregulation poses more competitive pressures, and competition forces you to
think about your customers.
What steps should corporate IT
users take to realize the benefits of datawarehousing?
As far as what companies should be doing, I think that they should be trying to begin with
installing a culture in the organization which enables them to take decisions based upon
facts.
When will datawarehousing become
mission critical, both internationally and in India?
Internationally, I still haven’t seen datawarehousing becoming mission critical. Nobody,
as yet, considers datawarehousing as being really mission critical, which means that
people have not started demanding HAS for their datawarehousing applications. Since it has
not yet happened internationally, I think that it will be sometime before it becomes
mission critical in India.
Which is the ideal OS/Platform for
datawarehousing?
Microsoft’s Windows NT is a viable choice for datamarts, which are small, focused
datawarehouses. For truly large enterprises, which require a high degree of scalability,
and things like reliability, high availability and advanced manageability, Unix is
probably the best choice. But it will change with time, as Windows NT becomes more capable
every day, and better at handling large volumes of data.
HP is using its Foundation Tools to
link Unix and Windows NT. Does this spell anything important for datawarehousing? Do you
feel that HP’s datawarehousing solutions will now enable users to set up small datamarts
with Windows NT, and then, as their data store grows, move on to Unix solutions at the
high-end, and then seamlessly integrate the two?
I’m not exactly an expert in that area, but the logical way is to let Windows NT in the
enterprise run datamarts, which are fed by a datawarehouse which runs Unix. Integration
between the two would then be critical, and this is where our Foundation Tools for
integration can step in.
Voltaire in Diatribe du docteur
Akakia is said to have apologized for making the statement, "although it seems to
contradict reality, we must trust our computations more than our good sense." Is
there any danger of people ignoring their common sense and trusting datawarehousing
results, resulting in their inevitable doom, as computer-enabled facts take over from
good-old common sense?
I think that it is an issue of making decisions based upon facts rather than hunches.
Companies have been managed on some people’s intuitions, judgements, or hunches. There are
some companies which are managed in a very autocratic manner. What datawarehousing says is
that sometimes, there are hidden relationships between facts which we are not very easily
aware of. Datamining tools can help us unearth these hidden facts and the relationships
between them. Companies can leverage on that and improve their revenues and make better
profits. Of course, it is always useful to reason things out yourself, and then use
datawarehousing applications to help you test things out with hard facts.
