SAP HANA Effect Episode 10

SAP HANA EFFECT
Title: Episode 10 - Real-Time Transformation at Dell (Duration: 18:39)
Publish Date: March 23, 2015
Description: Bart Crider shares how Dell has revolutionized their sales
and planning processes with real-time data for their entire sales force.
THIS IS A RUSH TRANSCRIPT. THIS COPY MAY NOT BE IN ITS FINAL
FORM AND MAY BE UPDATED.
JEFF:
Welcome to The HANA Effect. I’m your host, Jeff
Word, from SAP. Each week, we bring listeners the
real stories of how companies are taking advantage
of real-time computing to transform their
organizations and let them share the lessons they’ve
learned along their journey. Welcome to another
edition of The HANA Effect. I’m your host, Jeff Word,
and I’m here today with Bart Crider from Dell.
Welcome, Bart.
BART:
Ah, thank you.
JEFF:
Today, we’re going to talk a little bit about what Dell
has been doing with SAP HANA and their use case.
But before that, Bart why don’t you kind of introduce
yourself and what your role is at Dell.
BART:
Sure. Again, my name is Bart Crider, and I run our
Enterprise Business Intelligence department within It. My
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 1 organization has responsibility for any project delivery,
architecture, engineering around business intelligence and
analytics or enterprise performance management for Dell.
JEFF:
You’re actually one of the guys at Dell that has to
make this stuff work for your own internal purposes.
BART:
Yep. Now, I’m one of the back office geeks.
JEFF:
Awesome. We love those guys. So, Bart, why don’t
you give us a little idea of what you’ve been doing
with HANA and kind of describe your use case a little
bit for us.
BART:
For many years, Dell has kind of focused on an
architecture where we had one centralized enterprise data
warehouse and started to recognize that the nature of data
was changing. The technology was changing. And more
and more organizations from a best practice perspective
were moving more towards kind of an ecosystem of
technology, so that you applied the right technology to the
right type of data vs. trying to have this one solution that
fits all needs. We recognized that there was a big
movement toward capabilities from an in-memory
perspective. And Dell’s got a strong go-to-market
relationship with SAP and so, started to look at how we
could take SAP HANA and position that in kind of that inmemory capability as it pertains to our enterprise data
warehouse and BI ecosystem. And so, we partnered with
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 2 SAP a couple of years ago to kind of bring HANA in and
kind of prove out the metal, so to speak, and learn more
about it and how it would fit into what our enterprise
standards and objectives were, look at performance
perspective and compression perspective, and the ease of
putting BI on top of it. We also have a very big self-service
focus here and wanted to say, hey, how agile was this
platform to develop in and how much could actually put
into the hands of the end users in order to help them to be
self-sufficient in kind of meeting their reporting needs?
That project was very successful in helping us kind of learn
more about HANA and how to appropriately position it in
the different use cases that may come up from our
business partners. And as luck would have it or it’s maybe
a little bit of planning, we had a situation where we have a
legacy set of systems that support a lot of agent
performance and business attainment metrics, right?
These are the systems that the sales guys use. And you
know, Dell’s got a very complicated and matrixed oriented
sales organization and so you have to do a lot overlays and
drill downs to kind of look at how one agent is performing
to the next, or what the product mix may be for a
company or how a particular customer is buying across our
product line. The current solution, quite honestly, wasn’t
scaling, right? You know, sales guys would take a drill
down into looking at their data and, if the data came back,
it came back in minutes and not seconds and, obviously,
for a sales organization that’s not acceptable. Sales teams
need to be talking and working with our customers, not
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 3 waiting on specific queries to come back around their
information, so that they can make decisions. And so, we
work back with the business to say, hey, we think this is
the perfect use case for an in-memory technology and that
we have been positioned to solve this problem with SAP
HANA.
And so, we partnered with them on a very short POC,
again, to kind of bring in some of the actual data that was
in these legacy systems, prove out the performance of it,
help them to understand the mechanics of it because we
have a very collaborative relationship with our business
where actually they do a lot of the reporting, and so it
needed to be something that from an end to end
perspective could work for both sides from a solution
standpoint. And at the end of that POC, they were all in
and said let’s go do this thing.
So, for all of last year, we migrated from that legacy
solution into SAP HANA and we actually used the
opportunity to introduce agile development methodologies
into the approach that we took from a program and project
execution perspective, so we set up three week sprints.
We established the scope for that sprint and we’re rolling
stuff out every three weeks where the business could
actually see the data that was moving from the legacy
environment to HANA. So, we had a series of reports that
we produced over the course of the year around how
sufficient was the pipeline, how’s the business attainment
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 4 doing? I can look at it from a segment perspective, all the
way down to an individual contributor, I can compare that
across teams. We have things where, hey, how well are
my trips turning actually into interactions from Dell from a
quote or order perspective? Hey, how effective are my
quotes turning into orders? And what’s going on from an
overall participation with my customer in terms of buying
across my product mix? We have tons of information and
metrics that we have built into HANA and then roll that to
the business and retire the legacy system.
The effort was so successful—it’s called our Business
Management Scorecard or BMS—that they actually want to
take what we did for sales and kind of role the same
construct out across the other business units. It’s part of
what the business initiative was driving was to simplify and
standardize around a global set of KPIs and bring a lot of
the business and the processing back to the data vs. the
temptation to have to go push the data out to meet
individual use cases at a segment or function level.
So, that’s what we’ll be doing this year is just taking that
and extending that across into other areas. You know, one
of the things we saw from an agile perspective with HANA
was very flexible to make changes and we were able to
react when business priorities changed mid-year and they
needed something different from a data structure
perspective or data set perspective; very easy to make
those changes in memory. So, that’s been a huge
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 5 advantage as well in kind of improving the overall velocity
of the team and how quickly we could make and adjust to
information needs of the business, so they can make
decisions or have insights into how the business was
performing.
JEFF:
Let’s go back and actually double-click on a couple
things that you mentioned there. The first was you
talked about the sprints and these three week
sprints and the project team. Can you give us an idea
how many people were kind of involved with this
from your team and from the business side of things
to make that project a reality?
BART:
So, I would say we probably had on my team it was about
10-12 people. And then on the business side, it was
probably a core set of 4-6 and we kind of laid this out in
four main chunks.
We had one where did we have data gaps? And we needed
to move either specific data that they had sitting outside of
our ecosystem back in. Or, second, hey I move it and I’ve
got to do the business rules. And then I needed to move
the data into HANA and then I needed to go build the
reports. And so we had different workloads going on
around each one of those areas and across each one of the
sprints that my team or the business team was working
on. And so, my team was responsible for all of the data
ingestion, re-instantiating the business rules inside of our
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 6 ecosystem, pushing information out into HANA and
creating the analytical and attribute views. The business
actually owned the calculation views within HANA and then
the publish creating the universes and the reports. And so,
that’s kind of where we had the partnership. And so, we
would identify certain scope items across each sprints
where I may be doing, hey, in this area bringing new data
in to solve gaps while you’re building out the calculation
views for this. And the next sprint’s hey, now I’m pushing
that data into HANA while they’re building the report. So,
we had a real cadence that we went through over the
course of the year.
JEFF:
So, that sounds like you guys were just fast and
furious for quite a while there. What was kind of the
split between the effort involved with the data
quality, data sourcing, data provisioning side of it,
vs. the actual technical hands on with HANA to get it
going the way you needed it?
BART:
The bulk of the effort was all around building out the
models within HANA. I mean that’s where the IP is and
secret sauce that makes the performance so great. If you
don’t take advantage and understand how to build the
models, you can make any system perform slowly, and we
definitely didn’t want to do that with HANA. So, we actually
had partnered with SAP to bring specialists to come in to
invest in the models themselves, so that we knew that the
different engines inside HANA we’d structured the data
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 7 appropriately to optimize performance for those. So, I
think, by and large, that’s where the bulk of the effort was.
Moving the data in and using data services to push into
HANA was pretty straightforward, but I think the bulk of it
was really optimizing those models to get the performance
that we want and then, secondarily, creating the
calculation and creating the reports to push out to the
business.
JEFF:
Can you give us a kind of general idea of what the
size and scope of the data set that you’re working
with is?
BART:
We brought in a year’s worth, or five quarters, of our order
information from a facts perspective and then brought a
customer information, product information, CRM
information, sales teams-sales territory information and
quote all together and it’s probably several terabytes of
data that we put into HANA.
JEFF:
How many of those source systems are actually SAP?
BART:
Zero. Yeah, one of the unique things about Dell is you
know there’s kind of three main areas that we see HANA
positioned when you look at how SAP talks to customers.
And one of them is when you’re not a SAP ERP shop, which
we are not, and so we’ve positioned it truly for that
business intelligence and analytics capability in-memory,
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 8 and so none of the source systems that we get in for that
are coming from SAP back office systems.
JEFF:
So, this is a standalone, agile data mart with HANA
and analytics on top of that pulling in data from lots
and lots of systems, lots of data, and none of it
coming from an SAP environment?
BART:
Correct.
JEFF:
That’s pretty awesome. Let me ask you this: what
kind of problems did you have during the project?
BART:
Initially, we’re a pretty mature enterprise data warehouse
organization and you know being a large, global company,
managing mixed workloads is something that we’re used
to having a lot of administration and tools around. I think
one of the things that we have seen—and it’s not
unexpected given where HANA is in it maturity curve—you
don’t have a lot of the same capabilities from a HANA
perspective in terms of being able to manage and protect
yourself from a bad query bringing down the environment
or being able to manage how much capacity you’re going
to throw towards data loading vs. serving up the reports.
So, you have to have a little bit more hands on
administration support to do that. Again, we’ve actually
partnered with SAP who provides database operations and
oversees that environment to be sure that we’re doing that
correctly. I think the other thing that you have to be
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 9 mindful of is that it’s not just about HANA. It’s about the
whole SAP ecosystem that you have to build on top of
HANA and having that all lined up from a versioning
perspective, from a connection perspective, you have to
pay a lot of attention to configuration management to be
sure that the right version of HANA is compatible with the
version of business objects that you’re using or the mare
that you’re using or design studio that you’re using or
crystal reports that you’re using and you’ve got to kind of
keep that in sync.
Initially, there was a lot of patching going on, right, that
would actually break those connections. It’s smoothed out
a bit. It’s gotten better over time. But it’s still something
that you have to pay attention to. I think the third area
would be capacity planning. I think this is where we did a
good job up front, but the good news story of you guys
continuing to invest in new capabilities of HANA has
continued to kind of stress that capacity. We actually
partnered with SAP to say, hey, help us think through how
big of an environment we need. Because the temptation is,
oh, just you know, how much data do you want to put in
memory? But it’s also how many users you’re going to put
on the system. Because each one of those as they interact
with the data in HANA is going to consume memory.
So, you had to have a balance of, hey, we’re going to have
14,000-15,000 users on this. At the same time, we’re
going to be loading up multiple terabytes of data in this.
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 10 How much memory is needed to support all of that? When
you start adding on a lot of the excess services that have
happened since we kind of did that initial capacity
planning, whether it’s Lemari server or some of the others,
all of that goes and takes advantage of that same memory
too. So, I think having good capacity planning up front and
partnering with SAP in terms of understanding what new
products are going to continue to be put into the HANA
Suite is also critical.
JEFF:
So, that’s phenomenal advice. Something else I
wanted to check about. Sometimes people are
surprised in bad ways; sometimes in good ways.
What were some of good surprises that you guys
encountered on this journey?
BART:
We saw a lot more compression than we had expected
going into it and working back with SAP, we kind of
expected you know 1-5 was kind of the benchmark with
which to size. And we’ve actually seen 1-15, 1-20, which
we’ve loaded a lot of data input. So, we’ve been able to
put a lot more data into HANA than we expected; that was
a pleasant surprise. What met kind of our expectations,
but it was still pleasant is the performance. And we’ve
seen 60 times improvement for queries in HANA than you
would run on our traditional enterprise data warehouse.
And so, the speed of it has kind of been to the business a
pleasant surprise because it’s one of those things, yeah
yeah yeah, I believe it when I see it. You’re like, oh, you
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 11 guys were telling the truth; this is a whole lot faster than I
have today.
And I think, thirdly, was really having the business
embrace the agility of the platform and seeing the power
of it and how they could take advantage of it to make
publishing reports a whole lot more smooth and
streamlined than it is today. You know, when you have
both the business and IT rallied around a technology
believe it’s the right way to go, that makes all the
difference in the world. And so, having their participation
and getting to that understanding was key as well, and I
think it was a benefit to the success of the program.
JEFF:
You know, the speed is the sexy part. But there’s a
real business benefit behind that. Have you guys
been able to quantify at all the actual business
benefit of it this new system?
BART:
We have in probably some metrics that may be a little bit
different than you saw before. Actually, the 60 times
improvement in speed, obviously, from a sales perspective
is a business benefit. They also used this, this was part of
an effort to globalize and standardize a set of metrics. And
so, they saw a 40% reduction and kind of non-standard
metrics, as they rolled this program out. And so, you’ve
got some very tangible numbers. They don’t necessarily
equate an X million dollars or an X hundred thousand
dollars, but from a BI competency perspective, those are
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 12 very meaningful metrics to show how you’re optimizing
your overall BI spend and leveraging effectively across the
company.
JEFF:
Did you ever get any interesting reactions from the
users, other than wow it’s fast?
BART:
When can I have this? When can I have that? More, more,
more, more.
JEFF:
That is almost universal response is once they see it
for one thing, they immediately go, I want it for this,
this, this, and this, and I want it next week too.
BART:
Yeah. And that’s where you have to be smart about it,
right? Because the temptation that you run into, I think,
with any technology that you have initial success around is
to continue to do more and more and more with it. But, at
some point, if you don’t have any governance around that,
you can cripple that solution no matter what technology
it’s in. And so, we’ve had to put a lot of governance and
due diligence around wanting to cram more and more and
more and more data into HANA and get away from its core
purpose from a sales performance management system
perspective. And so, that’s something that’s been critical
too, to you know not fall in love with it. And then all of a
sudden, you’ve overused it and over taxed it and it doesn’t
really meet your needs anymore. So, I think that’s critical
too, to have the right governance structure in place to be
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 13 sure it says, hey, this is what we purchased it for. This is
its sweet spot. This is what it’s meant to go do. It’s
succeeding at that. Any decision to move away from that
has to have a very well thought out governance and
decision making process before we decide to do it.
JEFF:
Some people might not know, but Dell is actually a
certified and very good go to market partner with
SAP specifically on the HANA hardware. A lot of
questions I get asked, especially since you guys are
using this yourselves to help your business run
simpler and more efficiently. How do you guys work
and feed the knowledge that you’ve gained from
your internal usage out to the field teams that are
actually out there selling and installing and
supporting this at your customers?
BART:
Yeah. So, we have a very active engagement with our
engineering team that works with SAP on the certified
configuration, as well as with our sales organization in
terms of selling it out in the field and our service
organization because we have a service component around
SAP HANA as well. And so, for example, when it was Dell
World, we spoke about that. We’ve spoken at it at
Sapphire alongside the Dell booth there. I was actually just
on a call earlier this week where, again, they want to use
the success story we’ve had around the BMS to reeducate
the sales teams as they sell HANA into the market. And so,
it is very much an active discussion that goes on
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 14 frequently within Dell to keep all the parties involved
engaged in understanding the Dell on Dell story around
HANA and how we can share that with other customers
and show them the value of it.
JEFF:
I really love that virtuous cycle of getting your
experiences out to customers, but then also
customers giving you guys new ideas and things to
do as well. So, I do have to ask specifically about
that. Can you give us an idea of what this multiterabyte system, what kind of hardware landscape
are you guys running it on?
BART:
So, we’ve got a five node cluster—four hot, one standby—
2-terabyte scale-out environment that we are running the
sales BMS on.
JEFF:
So, each one of those five is a half terabyte
machine?
BART:
Each one of those five is half terabyte, runs on Dell 920’s.
JEFF:
920’s, nice. And what kind of advice would you give
to new customers who are considering going down
this path of an agile data mart with HANA?
BART:
Know what you’re doing with modeling; invest in that skill
set because that’s the heartbeat of HANA and to make it
perform and then also understand that you need to
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 15 manage what you know, don’t be cavalier in what queries
you execute, especially in a production environment
because, again, you don’t have all the work management
controls to be able to control the impact of some of those
changes. So, I think, again, from an administration
perspective, you have to take a little bit more hands on
and oriented approach to insure overall environment
stability.
We do quarterly health checks with the remote database
operations team from SAP that partners with us. So, they
tell us, hey, here’s where things aren’t performing as well
as we would expect. They look across the engine. Here
might be some hotspots you want to go invest. And it’s
uncovered some areas that we may be able to go optimize
around additional partitioning, etc. But I think to anyone
starting out, go make that investment in understanding
how to go model in HANA because that’s the key to it all.
JEFF:
What other aspects of HANA and the capabilities of
HANA are you guys looking to explore the next
couple of years?
BART:
You know, we’ve started to explore the other tools that
interact with HANA. Part of the beauty of it is that with
data discover tools like Lamara or something that you can
implement that’s more application oriented like design
studio that you can take for mobile, part of the beauty of it
the data all resides in HANA and it’s just the queries that
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 16 get saved. So, it minimized data movement immensely,
whereas opposed with other tools, they may be pulling
that down to the client or in other areas. And so, I think
the advantage of HANA and the ecosystem from a BI
capability perspective that’s being built around it is
fantastic because it really leverages the power of the data
in memory and avoids a lot of unnecessary data
movement. So, I think as SAP continues to build out
capabilities in those tools, it only enhances what HANA
brings to the table.
JEFF:
Bart, that is phenomenal information. I know
customers are going to get a lot of value out of what
you guys have done and using you guys as a great
example on how to do a specifically non-SAP HANA
environment using lots of non-SAP data and setting
it up as an agile data mart. We’re thrilled that you
guys are such wonderful partners of ours. We’re
thrilled that, more importantly, as an SAP HANA
customer, you guys are getting massive value out of
it and see a lot of new opportunities to really
simplify your business with that. For everybody else
that’s listening, I want to thank you, again, for
joining us on The HANA Effect, and you guys can get
more information, obviously, about SAP HANA on
hana.sap.com. And we’ll actually post some links for
this to the Dell page, where you can get more
information about how they’re using HANA and can
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 17 provide you with HANA services. So, thanks a lot,
Bart.
BART:
Thank you. Enjoyed the conversation.
HANA Effect Episode 10 – Real-­‐Time Transformation at Dell 18