Enabling the Managed Cloud: Insights from Our First Off Premise Migration to Virtustream
It’s the holiday season and that time of year where we get together with friends, family and loved ones. Big gatherings are common which are fun for some, stressful for others. My family has grown beyond the dining room in my Mom’s house, so we now find ourselves having the big family holiday dinner quite literally in the play room above her garage. Yes, we carry all of the holiday dinner out of the kitchen, up the stairs, through a hallway, and into this room, because it’s the one big room in the house where we can all comfortably sit and eat. So Mom gets a little stressed, but luckily she has a nice, empathetic family. Inefficiency be damned, we optimize for the moment.
IT executives face a similar dilemma in managing day-to-day data center operations, but their customers and user base are a touch less empathetic. Fortunately, we now have the ability to expand off premises, so we don’t have to buy that big expensive house anymore and can pay for capacity on demand. But we still struggle with how to optimize those expansion capabilities and manage our TCO of our enterprise IT infrastructure assets.
Recently, as part of our hybrid cloud strategy, Dell IT successfully deployed a mission-critical application, SAP Sales & Operational Planning (S&OP), onto the Virtustream cloud. We had talked about our plans to adopt Virtustream’s Infrastructure-As-A-Service (IaaS) offerings a few months ago in our Moving Toward a Seamless Cloud blog, but now we’re there. We’ve opened the door for the ultimate goal of dynamic workload management for mission-critical enterprise applications across on-prem and off-prem assets, optimizing costs and resources with zero impact to our growing business. We still have a journey ahead to reach this goal, but the first step of getting a real production system live on the Virtustream Enterprise Cloud was an enormous one—a true milestone. We wanted to share this experience so others can gain from our learnings.
Why Go Off-Premise?
Over the past several years, EMC IT (now Dell IT) has implemented an SAP ERP solution leveraging virtualized Vblock converged infrastructure to accommodate a growing number of applications and workload types. However, as our applications and business use cases continue to evolve, we face infrastructure capacity expansion challenges. Rather than investing in ever-increasing IT infrastructure on-prem, we opted to methodically migrate some of our SAP workload off-prem to optimize efficiencies across our application portfolio, as well as freeing up on-prem capacity.
A key part of our strategy is to use Virtustream’s unique software, xStream and MicroVM (μVM) cloud resource management technology, to accurately measure our IaaS utilization and bill business units for the services they use—achieving a so-called “pay-by-the-drip” model.
To launch our IaaS model, we initially focused on applications that were ‘loosely coupled’ in nature, meaning low-volume, batch-based file transfers to support the data translation and manageable user volumes. Our SAP S&OP application is utilized by some 200 global Sales Operations team members to forecast demand, manufacturing and procurement processes. Being a planning application, utilization peaks for a couple days monthly, increasing HANA database demand to a little more than one terabyte versus the average utilization of less than 100 gigabytes. The app is fed daily from multiple data sources, including master data and information on products, locations, shipments, bookings and billing. So not only did S&OP meet our loosely coupled criteria, but its fluctuating infrastructure demand allowed us to take full advantage of xStream’s pay-by-the-drip provision.
Virtustream offers two options for such managed cloud projects—one in which it supplies the IaaS and a second where it also provides customers with Managed-Service-Provider-type (MSP) BASIS/DBA support along with IaaS. We chose to focus more exclusively on the IaaS model and manage the software / HANA installations and support ourselves, taking advantage of our home-grown, top-class HANA BASIS/DBA and development skills. We also felt more comfortable with a hands-on approach and wanted to be able to maintain our close collaboration with the business.
The process began with a couple of parallel paths. While Virtustream’s on-boarding team worked with our network architects to plan and implement network connectivity, they also partnered with our SAP team to determine our IaaS capacity needs on their cloud.
To establish network connectivity, we started by using a VPN (Virtual Private Network) tunnel to get the project off the ground, with the expectation of deploying a more standard recommended MPLS (Multi-Protocol Label Switching) solution to provide more bandwidth for broader scale. Throughout testing, however, we determined the VPN connection had ample capacity for this particular S&OP application, and thus the MPLS setup was not critical path. So while we’re now live with VPN, we are continuing to pursue the MPLS deployment in lieu of future migration projects.
We did face a few challenges in adequately sizing the target off-prem environment. Virtustream’s analysis of or our ‘as is’ footprint and workload was done as the S&OP application was still maturing within our roadmap. Therefore, we weren’t able to sufficiently factor in our (once monthly) peak utilization in aggregation with a forecasted application rollout plan. More specifically, we were about halfway through our global S&OP rollout when the analysis was performed, so we needed to interject our own resident knowledge of the app utilization patterns and implementation plans to right-size the off-prem platform.
The lesson learned is to know your app – present and future. While current workload patterns are important, be sure to consider roadmap and business implementation plans when planning your migration.
Testing is Key
We took a typical deployment approach of testing out the app in a sandbox environment and then methodically worked through our non-production landscape, to our Performance Testing instance and then into Production. While we had business users functionally testing the applications, the SAP team leveraged our in-house load testing tools to validate performance and scalability on the target platforms. Since the business had concerns about performance implications of running a mission-critical application off-prem, load testing was critical. By using our own load testing tools, we were able to provide ‘before and after’ metrics from within our own network.
The SAP S&OP application architecture is unique compared to traditional SAP apps: There is no application/Netweaver layer on top of the database, but rather the entire app is run within and accessed via the HANA database itself. Thus, the technical process of migrating S&OP was a relatively simple database migration, i.e., backup and restore. After our BASIS team installed the shell HANA instance(s) on the IaaS hosts provided by Virtustream, they performed a database restore off of the latest on-prem backup. The whole Production migration process was a mere couple hours with minimal downtime.
An important feature that made the move seamless for our users was the fact that launching the app in the new environment didn’t require changing the hostnames. Virtustream used our internal hostnames and aliases, so there was zero impact to the user access paths and they saw no change whatsoever.
The benefits of the move, however, are clear. With a consumption-based model, we only pay for the capacity we use. And paying by the drip allows us to shift our IT investments from capital expenses (CAPEX) to operational expenses (OPEX), a more strategic incremental approach to funding infrastructure that eliminates lump-sum capital budget complexities and accompanying tax requirements.
We have also freed up significant high-density compute capacity—in excess of three terabytes of HANA memory—that we can now re-allocate to other on-prem applications. And, we are working with Virtustream to evaluate potential use cases to shift additional workloads off-prem, of course factoring in future business and IT roadmaps.
Stay tuned for further insights on our managed cloud journey, as we continue our efforts to create seamless workload management of our mission-critical enterprise applications across on-prem and off-prem assets.