|
“There are two paths to get to the Cloud—the evolutionary path and the revolutionary path”
Sally Stevens, VP, PG Enterprise Platform Marketing,
Dell Inc. and Murali Gopalan, CIO, Sr. VP, UST Global talked to Prashant
L Rao about virtualization, server trends, why the public Cloud is too expensive
on a per VM basis in India and more. S Sridhar, Director - Marketing,
India Relationship, Dell India, and Rinosh Jacob Kurian, Enterprise Architect,
UST Global chipped in.
What is Dell doing on the server virtualization front?
 Sally:
Virtualization has been great but customers are looking for more of an RoI for
which you need automation and transformation. There are two paths to get to
the Cloud—the evolutionary path and the revolutionary path with our DCS
platforms with different hardware and cheap compute nodes. In the next three
to five years, traditional server form factors are going to change. The revolutionary
is going to transition into some of the evolutionary form factors.
How are things going in India in the server market and
on other fronts?
Sridhar: In Q1 we were the joint #2 in servers. The
revenue share has grown faster than the unit share. Almost every single acquisition
that we have made is available in India. Most of them had development centers
in India and we are leveraging these resources and making them available to
our customers. We have about 600 engineers at our R&D center in Bangalore
and the design for our next generation of servers is being done there today.
A lot of customer PoCs happen at the center.
Have you hit $2 bn in turnover in India yet?
Sridhar: On a sequential basis, we have crossed $1.5
bn and we may hit $2 bn in revenues in 12-15 months time. For a company that
was $1 bn just 18 months back, that’s a significant accomplishment. We
are starting to export stuff from our factory in Chennai to the Middle East
and it’s going to expand further.
UST Global is a Dell customer. How did you get started
on your journey to virtualization?
Murali: We have developed something called the Apple
Tree, which is an innovation lab. We have large masses of people who want to
try out new stuff but, as our's is an IT services environment, it’s a highly
regulated one. We have entered into agreements with several vendors for Apple
Tree where you can come and try out new stuff. One of the limitations of this
model has been that it is confined to a physical location. Rinosh had the idea
of using virtualization to allow access from anywhere. The idea was to have
training licenses in a virtual network that could be extended to anywhere in
the world.
Rinosh: We have projected a capacity requirement of
a thousand VMs to be created this year. We wanted a high level of efficiency
as our data centers were designed about six years back and we wanted to stretch
out that investment in the power and cooling systems etc. We evaluated various
products on power efficiency and compactness in addition to raw horsepower.
Our CEO requested the CIO to look into processor utilization etc. We partnered
with Dell to get a hands on feel before the big purchase. We had some Dell hardware
procured for another purpose and we redeployed that and realized that we could
safely run 40-50 virtual machines on a four socket box. Today we have 480 physical
servers. By going in for a thousand VMs we are effectively doubling our server
count. We have been facing difficulties in patch management etc. Going forward,
adding physical servers was clearly not a scalable model. We were introduced
to Dell solutions such as VIS that could help. We agreed that we would go for
that in a structured manner, probably in the next year. Dell’s ability
to manage the hardware from the rack perspective allows us to manage our data
centers in Chennai, Cochin and the US and the disaster recovery center in the
US from our primary data center in Trivandrum. If something’s switched
off, we need to bring it up remotely. We wanted a partner who was continuously
committed to improvement, had a plan and was willing to share that plan with
us.
Murali: The proposal that came to me was to spend
half a million to a million dollars to augment the data center. We decided to
explore the public Cloud but the cost per VM just didn’t make sense. As
soon as we decided to do the virtualization ourselves, the cost equation worked
out. Both public Cloud and on premises Cloud managed by a service provider didn’t
work out. It’s the management fee and taxes that are the killers right
now given the fact that the service providers don’t have the required scale
right now. Whereas, we have the advantage that we are in a STPI facility.
Rinosh: We had procured about 22 servers for virtualization
more than a year back. We had started with a different virtualization software
and now we have switched over to VMware. This new procurement is for the 1,000
VM capacity. Currently we have around 14 Dell servers that are completely virtualized.
It’s a mix of 2 and 4 socket servers.
At this point, it's mostly 2 socket but we are moving over to 4 socket as 2
socket machines have memory limitations. We have come to understand that you
are more likely to run into bottlenecks on memory than on compute as memory
isn’t growing at the same rate as compute.
We will have to cut over to public Cloud at some point in time but today it’s
not there in India. In the US, you can rely on the Cloud as the infrastructure
there is good whereas India relies on US servers for its complete Internet requirement
and the last mile in our country is typically not that good. It will take us
a few years to sort that out. In a few years, we would have to leverage the
public Cloud as the economies of scale would become a leverage for a service
provider by then. Today, we can do it in-house at a lower cost than any service
provider in India.
Murali: Their model is optimized for the US and the
design is over engineered.
What kind of demand do you expect to see once you go through
with this deployment?
 Murali:
When airfares were liberalized no one could have predicted what the demand would
be. Today, it takes six weeks to provision a server. We are moving to a model
where we will be able to do that in 15 minutes and we don’t know how it
will translate into our business.
How important are partnerships with customers for Dell?
Sally: We want to have customer inspired designs.
You can do things in hardware but it’s about management these days.
What technology trends do you see in the server market?
Sally: The growth of blades has slowed down and you
are starting to see the emergence of what we call shared infrastructure. Blade
adoption continues to grow, however. There are advantages to blades like the
cabling being simpler, more intelligent management—you can preset the personality
of what you want the blade to be.
Sridhar: Blade sales in India account for 14-15% of
the overall server market.
At what point do you see shared infrastructure becoming
a part of mainstream computing?
Sally: It will be a while before shared infrastructure
is adopted in the private Cloud. However, it is being adopted in large volume,
large scale deployments.
Shared infrastructure is like a blade but you don’t have some of the components
and the infrastructure is shared more and it’s better in terms of power
and cooling. However, it requires different stacks of software and redundancy
is handled at the application level.
People who buy this are buying it in thousands of units. We started with DCS
and we have started to bring that down with our PowerEdge C series but it’s
still for the large volume segment.
Tell us about something innovative that Dell's done on
the green computing front?
We announced our FreshAir initiative where we are allowing our servers through
tweaking of configurations to run at 40 degrees Celsius.
What's your take on in-memory computing?
Murali: Two years ago, a worldwide survey of CIOs
said that business analytics was going to be the fastest growth area.
We deployed IBM Cognos TM1 for planning and budgeting. At the end of the day,
if you don’t deploy it well, in-memory’s going to boomerang on you.
Having said that, in-memory is the future as the data overload is happening
to an extent that you need in-memory compute for decision making.
We have procured the hardware about 15 days ago. The project is in full swing
and we hope to launch it in the second week of September.
|
