Microsoft has made a series of cloud power plays that have ramped up in frequency and ferocity since the turn of the year. Under the umbrella of their cloud push are not only new features to extend Azure, but also added virtualization-related services, including this week’s introduction of System Center 2012, which targets private cloud-building.
To help provide a case study to back the use of System Center 2012 and their own flavor of virtualization technology on the massive scale (and for mission-critical operations) the Redmond giant announced that American mega-retailer Target is running a number of business-critical workloads for all of its stores across 15,000 virtual machines.
According to Microsoft, Target has virtualized nearly every aspect of its business-critical operations, including inventory, point of sale, supply chain management, asset protection, in-store digital input and other processes using Windows Server 3008 R2 Hyper-V and Microsoft System Center.
The feat to merge together a distributed infrastructure with over 300,000 endpoints, which include not only servers (15,000) and workstations (29,000), but also registers (52,000), mobile devices (70,000) and even a range of kiosks at different points throughout Targets 1,755 locations is no small one. The move will eliminate 8,650 servers by 2012 and will adhere to a “two servers per store” policy.
Fritz DeBrine, who heads up Target’s Server Technology and Enterprise Storage division noted that the prime motivation behind the two-server policy boiled down to cost as prior to building their virtualized army, there were several servers, most often seven, dedicated to different tasks in each store. DeBrine said that the power, maintenance and refresh costs were quite hefty.
A Targeted Tale of Virtualization
Although Target didn’t begin the virtualization process until 2006, general server sprawl following rapid expansion and an increasing complex, distributed system were management challenges that the company first identified in 2004. At that time their first idea was to replace outdated servers that supported their pharmacy operations, which ran on IBM AIX.
Brad Thompson, Target’s Director of Infrastructure Engineering claimed that at this point in 2004, “the hardware we had been using for the pharmacy solution was no longer available. At the same time, we didn’t want to replace a lot of the hardware as that would cost us millions of dollars. By this point, virtualization had become a viable alternative to simply deploying more servers and would allow us to make better use of existing server capacity and reduce infrastructure management.”
The migration toward virtualization was not something that happened overnight, certainly, but the process has been rolling since 2006 when IT leaders at Target decided to roll with Microsoft Virtual Server. Like other large enterprises that make the decision to examine virtualization routes, DeBrine stated that the goal from the outset was to reduce the number of servers. He explained that before the virtualization process the company operated under a buy a server when you build a new application model—one that is cost and operationally inefficient.
DeBrine claims that they carefully weighed their options at this early point, examining the offerings of VMware closely before finally settling on Microsoft. He notes that at that time, the first VM they deployed was “a SUSE Linux instance running our pharmacy application. Things were runningwell and created three additional workloads on MSVS over the next 18 months” and kept adding more applications to the mix.
The art of piling on additional workloads, however led to a performance bottleneck, which in turn led to the company’s decision to deploy Hyper-V in a span of 45 days. DeBrine noted that since that time Hyper-V has become central to their strategy going forward as they seek to meet their 8,650 server-elimination goals.
The final piece of the puzzle didn’t fall into place until 2010 when Target started a pilot project to migrate its existing workloads to Hyper-V in ten of its highest-volume locations during the busiest time of year—the holidays. This risky test proved successful for the last virtualization phase, which involved the purchase of Dell R710 servers for the hosts that ran Hyper-V and Dell MD1000 Storage units.
Vice President of Technology Services at Target, Jeff Mader, echoed DeBrine’s assessment that using Hyper-V allowed the company to reduce server footprint “without sacrificing the mission-critical application performance.”
The System at the Center
While in many ways Target’s story highlights the use of Hyper-V, it is also hoping that word about its capabilities to manage complex, distributed environments via System Center 2012 will be the main takeaway.
According to Microsoft’s Corporate VP, Brad Anderson, the move goes far beyond creating virtualization capability. In his words, “virtualization and server consolidation are important steps toward cloud computing, but its essential to have management tools that provide intelligence about how the apps themselves are doing, not just management of virtual machine black boxes.”
This extension of management capability, called Virtual Machine Manager was released as part of a beta program, which complements existing cloud-related features like the code-named “Concero” element, which allows application managers to deploy and manage applications on both their own internal and public clouds.”
The issue here is that there are a number of other solutions that do exactly the same thing that Microsoft’s dual-punch Hyper-V and System Center announcements target. Naturally, as an increasingly vocal presence in the virtualization and cloud space for large enterprise, the company hopes this type of solution will be accepted in the enterprise as a viable competitor to other possibilities that offer the same principles in practice from a deployment and management perspective.