Tuesday, December 20, 2011

SAP Runs VMware to Provision Virtual Machines to Support Complex Training Courses

Transcript of a BriefingsDirect podcast on how SAP uses VMware products to implement a private cloud that smooths out educational apps runtime requirements.

Listen to the podcast. Find it on iTunes/iPod. Download the transcript. Sponsor: VMware.

Dana Gardner: Hi, this is Dana Gardner, Principal Analyst at Interarbor Solutions, and you’re listening to BriefingsDirect.

Today, we present a sponsored podcast discussion on how worldwide enterprise applications leader SAP has designed and implemented a private cloud infrastructure to support an internal consulting and training program.

By standardizing on a VMware cloud platform, SAP has been able to slash provisioning times for multiple instances of its flagship application suite, as well as set the stage for wider adoption of cloud models. [Disclosure: VMware is a sponsor of Briefings Direct podcasts.]

Here to tell us about the technical and productivity benefits of private clouds, is Dr. Wolfgang Krips, the Senior Vice President of Global Infrastructure at SAP in Walldorf, Germany. Welcome to BriefingsDirect, Dr. Krips.

Krips: Thank you, Dana.

Gardner: Tell me about this particular use case. You've needed to provision a lot of your enterprise resource planning (ERP) applications and you've got people coming into learn about using them and implementing them. What is it about private cloud that made the most sense for you in this particular instance?

Krips: Expanding a bit on the use case, there is a specific challenge there. In the training business, people book their courses, and we know only on Friday evening who is attending on the course on Monday. So we have only a very short amount of time over the weekend to set up the systems. That was one of the big challenges that we had to solve.

The second challenge is that, at the same time, these systems become more and more mission critical. Customers are saying, "If the system isn't available during the course, I'm not willing to pay." Maybe the customer will rebook the course. Sometimes he doesn’t. That means that if the systems aren't available, we have an immediate revenue impact.

You can imagine that if we have to set up a couple of hundred, or potentially a couple of thousand, systems over the weekend, we need a high degree of automation to do that. In the past, we had homegrown scripts, and there was a lot of copying and stuff like that going on. We were looking into other technologies and opportunities to make life easier for us.

A couple of challenges were that the scripts and the automation that we had before were dependent on the specific hardware that we used, and we can't use the same hardware for each of the courses. We have different hardware platforms and we had to adopt all the scripts to various hardware platforms.

When we virtualized and used virtualization technology, we could make use of linked cloning technology, which allowed us to set up the systems much faster than the original copying that we did.

The second thing was that by introducing the virtualization layer, we became almost hardware independent, and that cut the effort in constructing or doing the specific automation significantly.

Gardner: When you decided that virtualization and private cloud would be the right answer, what did you need to do? What did you need to put in place and how difficult was it?

The important piece

Krips: Luckily, we already had some experience. The big thing in setting up the cloud is not getting, say, vSphere in place and the basic virtualization technology. It's the administration and making it available in self-service or the automation of the provisioning. That is the important piece, as most would have guessed.

We had some experience with the Lifecycle Manager and the Lab Manager before. So we said at that time because we did this last year, we set up a Lab Manager installation and worked with that to realize this kind of private cloud.

Gardner: For our listeners’ benefit, what sort of scale are we talking about here? How many virtual machines (VMs) did you have or do you have running?

Krips: In that specific cloud, typically we have between a couple of hundred and a couple of thousand VMs running. Overall, at SAP we're running more than 20,000 VMs. And, in fact, I have about 25 private cloud installations.

Gardner: What is it about this particular private-cloud installation that ended up being a proof of concept for you. Was this something that offered insights into other instances where clouds made more sense?

This cloud also gave us some hints on where we have to redesign the workloads so that they become more cloud usable.



Krips: One of the reasons ... is the kind of criticality that we have here. As I mentioned, this cloud has to work. If this goes down, it’s not like some kind of irrelevant test system is down -- or test system pool -- and we can take up another one. Potentially a lot of training courses are not happening. With respect to mission criticality, this cloud was essential.

The other thing that was very interesting is that, as I mentioned before, we have to replicate a lot of systems from a golden master image. The technology that one typically uses for that is network fencing. So we started off with courses that used network fencing.

One of the issues that we ran into is that there are a couple of courses where you can’t use network fencing, because the systems need to connect to common back-end systems. This cloud also gave us some hints on where we have to redesign the workloads so that they become more cloud usable. That’s why I think this cloud implementation was very specific and very important for us.

Gardner: Are there specific payoffs? I suppose there are in just the reduced time for provisioning and the ability to then automate and to use that common infrastructure. Any other thoughts about what the payoffs are when you can do a cloud like this?

Krips: The payoffs are that in the past we had only the weekend as a window to set this all up. A couple of things had us scratching our heads. One thing was, the amount of time that we needed with our traditional copying scripts was significant. We used almost the full weekend to set up the courses. There was really very little room if we needed to fix something. Now, with linked cloning, that time was cut significantly.

Pay for itself

The other thing was that the effort of maintaining the automation script was reduced, and I could deploy a significant amount of the resources to work on more innovative parts like redesigning the workloads and thinking about what could be next steps in automation. If you look at it, with all the tools we utilized, the “cloud implementation” will more or less pay for itself.

Gardner: We often hear similar requirements being applied to a test and development environment. Again, bursting is essential, management and automation can be of great benefit, and it’s mission critical. These are developers are making products. So does that make sense to you, and are some of your other clouds involved with the test and development side of the business as well?

Krips: As I mentioned before, we have 25 private-cloud installations, and in fact, most of them are with development. We also have cloud installations in the demo area. So if sales people are providing demos, there are certain landscapes or resource pools where we are instantiating demo systems.

Most of the VMs and the cloud resourcing pools are in the development area, and as you mentioned, there are a couple of things that are important to that. One is, as you said, that there is a burst demand, when people are doing testing, quality assurance, and things like that. Almost more important is that SAP wants to shorten the innovation cycles.

Internally, we've moved internally to an HR development model, where every six weeks development provides potentially a shippable release. It doesn’t mean that the release gets shipped, but we’re running through the whole process of developing something, testing it, and validating it. There is a demonstrable release available every six weeks.

Moving to the private cloud and doing this in self-service, today we can provision development systems within hours.



In the past, with a traditional model, if we were provisioning physical hardware, it took us about 30 days or so to provision a development system. Now, if you think about a development cycle of six weeks and you’re taking about nearly the same amount of time for provisioning the development system, you’ll see that there is a bit of a mismatch.

Moving to the private cloud and doing this in self-service, today we can provision development systems within hours.

Gardner: That’s what I hear from a number of organizations, and it's very impressive. When you had a choice of different suppliers, vendors, and professional services organizations, was there everything that led you specifically to VMware, and how has that worked out?

Krips: I can give you a fairly straightforward answer. At the time we started working with private cloud and private-cloud installations, VMware was the most advanced provider of that technology, and I'd argue that it is still today.

Gardner: How about security and management benefits? It seems to me that security might not be quite the same issue when it comes to the training instances, but it would be with development, having that source code in control, particularly if you’re doing distributed development. Are there aspects of the private-cloud benefits for security management that are attractive for you?

Very reluctant

Krips: Certainly. The whole topic of cloud, in general, and the notion that workloads can run anywhere in the rut, as it would be in a public cloud, it's certainly something where I personally would be very reluctant when it comes to critical development systems and the intellectual property (IP) that’s on there.

From our perspective, we wanted to have the advantages of cloud with respect to flexibility, provisioning speed, but we didn’t want to have more security headaches than we already had. That’s why we said, "Let's get our arms first around a private cloud."

Even today, our cloud strategy is hybrid cloud strategy, where we’re implementing certain workloads in the private clouds, and there would be certain other payloads that we will potentially be willing to put into a public cloud. Still, development systems would be in 99 percent off the cases on the list where we would be saying they go only in the private cloud.

Gardner: Is there something about a standardized approach to your cloud stack that makes that hybrid potential, when you’re ready to do it, when it's the right payload, something that you'll be pursuing? How does the infrastructure affect your decision about moving to hybrid?

Krips: That’s one of our biggest problems that we're having. Clearly, if one had a standard cloud interface like a vCloud interface, and it was the industry norm, that would be extremely helpful. The issue is that, as you can imagine, there are a couple of workloads that we also want to test in some other well known cloud rents. I'm having a bit of a headache over how to connect to multiple clouds.

For us, it's very important that we separate the user data and the desktop from the device.



That topic is still one of the things that we haven’t finally resolved. Because we have to choose. We basically have to unbolt one external cloud after the other, and everything is still an individual integration effort. Now, if a couple of interesting providers had a standardized cloud interface, it would be very nice for me.

Gardner: This is the last subject for today -- and I appreciate your time and input. A lot of folks that I speak to, when they’ve gained some experience with private cloud and hybrid cloud, start to think about other ways that they can exploit it, that will bring them productivity and technical benefits.

And moving more to the mobile tier, looking at the client, and thinking about delivering not only applications as services, or as terminal services, but thinking about delivering the entire desktop experience, more and more of it as a cloud service, seems to be appealing.

Any thoughts about what your experience and benefits with cloud might mean for your future vision around clients?

Krips: Dana, the thing is pretty clear. If you look at the strategy that SAP pursues, mobility is an integral part. We also think that not only that business process mobility is more important, but what we’re also seeing, and I mentioned that before, with the agility and development. So for instance, there are people who are working every couple of months in new teams. For us, it's very important that we separate the user data and the desktop from the device. We’re definitely pushing very strongly into the topic of desktop virtualization (VDI).

SaaS application

T
he big challenge that we’re currently having is that when you’re moving to VDI, you take everything that’s on the user's desktop today, then you make out of that more or less a software-as-a-service (SaaS) application. As you can imagine, if you’re doing that to development, and they are doing some complex development for the user interfaces or stuff like that, this puts certain challenges on the latency that you can have to the data center or the processing power that you need to have in the back-end.

From our side, we’re interested in technologies similar to that view, and where you can check out machines and still run on a VDI client, but leverage the administrative and provisioning advantages that you have through the cloud provisioning for virtual desktops. So it's a pretty interesting challenge.

We understand what kind of benefits we’re getting from the cloud operations, as I said, the center provisioning, application patching, improved license management, there are a lot of things that are very, very important to us and that we want to leverage.

On the other hand, we have to solve the issue that we’re not blowing the business case, because the processing power and the storage that you have at the end point is relatively cheap. If you move that one-to-one to the back end, we would have difficulties with the business case. That’s why we were so interested in VDI technologies that allowed us checking out an offline mode. That would allow us also to take care of all of our mobile users.

Gardner: If the past is any indication, the costs of computing go down. When there is more volume involved, perhaps with moving to VDI, we should see some significant price improvement there as well. So we’ll have to see on that?

There are a lot of things that are very, very important to us and that we want to leverage.



Krips: Yeah. But we’re confident that we can get the business case to work. Particularly for us, the VDI, the benefits, are very much in the kind of centralized provisioning. Just to give you an example, imagine how easy it would be if you’re doing desktop virtualization, to move from Windows 7 to Windows 8. You could basically flip a switch.

Gardner: Wouldn’t that be nice?

Krips: Yup.

Gardner: Thank you so much. We’ve been talking about how worldwide enterprise applications leader SAP has designed and implemented a VMware private cloud infrastructure to support an internal consulting and training program, and how that has led them to even bigger and better concepts around cloud and the business and technical benefits therein.

I'd like to thank our guest. We’ve been here with Dr. Wolfgang Krips, the Senior Vice President of Global Infrastructure at SAP.

Thank you so much, Dr. Krips.

Krips: Thank you, Dana.

Gardner: This is Dana Gardner, Principal Analyst at Interarbor Solutions. Thanks to our audience, and come back next time.

Listen to the podcast. Find it on iTunes/iPod. Download the transcript. Sponsor: VMware.

Transcript of a BriefingsDirect podcast on how SAP uses VMware products to implement a private cloud that smooths out educational apps runtime requirements. Copyright Interarbor Solutions, LLC, 2005-2011. All rights reserved.

You may also be interested in:

No comments:

Post a Comment