Arun Ramachandran, Country Manager, Data Computing Division, EMC India & SAARC gives an insight to Jasmine Desai on Greenplum’s analytics solution, and a perspective on where analytics is headed in the future.
How does Greenplum view anayltics unfolding in India?
Analytics is worthless unless it gives you actionable result. If organisations are only going to store data then it is an expensive investment. But, if they want to make money out of it, they really need to look at how to leverage the data in an intelligent manner. Indian customers are looking at these to make differentiators for themselves. For example, one of our customer wants to do analytics using the data they control and through this process they can differentiate, become transparent and more nimble.
Where data analytics need to be done on a large scale, operational reporting does not make sense. We
are more popular where data volumes are very large. We have a large base in telecommunications, Internet companies etc. Greenplum offers parallelised analytics here.
How is mobility changing the entire analytics domain?
The whole objective of mobility is what you do with the information in hand. The model should feed in certain information and perform a certain action. Mobility should help you to get to that action. Its just a transmission device. We do not expect CEOs to run a data model. We expect CEOs to get the result in such a form, that he can initiate action using such a device.
We have a collaboration platform called Chorus, a data scientist, a data engineer and line of business owner can communicate. The idea is to communicate an intent Vs insight Vs an action related by where it started.
Presently, what trends are impacting the analytics market?
The reality hitting the market is that you need to combine analytics with action to get results. Compliance and regulations are forcing lot of CIOs to maintain data online. How much of it is employed to add value to the business? If you have been a customer with a bank for five years and all your data is stored for this much time, how is that data impacting the banking business? The requirement for regulatory compliance is only going to increase and with it, the need to gain value from this data.
Where do you see intersection of Cloud and analytics happening?
There are specific restrictions associated with data sharing and that is a roadblock for organisations who want to do everything on the Cloud. However, there are customers who have deployed Greenplum on the Cloud. For example, doubleIQ in Australia actually runs this as a service. Also, there is big motion to go Cloud, so legislation will also have to support this transition.
When it comes analytics it has to be customisable. It is useless unless it is actionable. The only way for it to be actionable is to be governed by processes of the company itself. Thus, what might be actionable for one organisation might be useless for another. That is where the intersection has to work really well.
What are your views on current analytics market?
The analytic market has not expanded, but consolidated. In 2010, EMC acquired Greenplum, very shortly after that IBM announced acquisition of Netezza and later Teradata announced acquisition of Aster. There is a lot of visibility around analytics because that is what customers need today. Analytics is the ability to
run parallelised statistical, artificial intelligence model against all your
data (structured or unstructured). Organisations should be able to use the same data internally as well as externally.
What is your take on in-memory computing?
When it comes to operational real-time analytics, in-memory is very important because it is very fast. But the moment data volume grows larger, it becomes very difficult to manage. Everything is not meant to be in-memory. When the data grows up, you should be able to scale the same job and run at the same speed. It imposes whole bunch of restrictions and complications that the underlying technology must support.
Scale-out parallelism is what is needed in such a case. Through scale-out parallelism organisations can run a certain volume of data in two seconds and when the data volume grows it still runs in two seconds. It is not that the database is fast and capable, but the integration with statistics is such that you run a job and statistical program in parallel too. It is not a simple programme to run.
