High-Availability Architecture Design: Coles started by working with Australian–based Xello, a member of the Microsoft Partner Network that has Gold competencies in virtualisation, systems management, and desktop. The companies have worked together on other strategic technology projects since 2008.
Over an 18-month period, the companies designed the solution architecture to be implemented at each store, which consists of two-node clusters running Windows Server 2008 R2 SP1 with Hyper-V—a commonly used high-availability architecture (see Figure 1). Coles uses two virtual server hosts in each of its supermarket locations: one serves as the application server, the other as the point-of-sale server. The application server runs applications that help keep the stores operating smoothly, including file and print services. In addition, the application servers run propriety software, such as applications for automated replenishment and inventory management, meat and deli scale management, and back-of-store functions. The point-of-sale server hosts the proprietary software that the stores rely on to process more than 11 million customer transactions each week, including electronic funds transfer, gift cards, and loyalty transactions. The two physical hosts, which run on Fujitsu Primergy RX300 S6 servers, are connected by a fully-redundant storage area network (SAN) that runs on a Fujitsu Externus DX60 storage server.
Coles uses Cluster Shared Volumes in its failover clusters, a feature in Windows Server 2008 R2 that enables virtual machines to access the same virtual hard disk files. At the same time, virtual machines in the cluster can move from node to node during a failover process independently of one another, without causing other virtual machines on the same hard disk to fail over.
The company also takes advantage of Dynamic Memory, an enhancement to Hyper-V that was introduced in Windows Server 2008 R2 SP1, to further enable high availability. By using Dynamic Memory, Coles can pool memory on a host machine and dynamically distribute it to other virtual machines in the cluster as necessary. The memory is automatically added or removed, based on server workloads, without service interruption, so that Coles can make efficient use of all available physical memory.
After designing the high-availability architecture with Xello, Coles also worked with Fujitsu to run pilot deployments and test the virtualisation solution in various store settings. In April 2011, after a successful pilot phase, Coles started rolling out the solution with Hyper-V to individual supermarket locations. By October 2011, the company had implemented the high-availability virtualisation solution to 300 stores and plans to have a total of 550 stores completed by the end of 2011. It will continue its deployment efforts at the remaining 200 supermarket locations throughout 2012. When the project is complete, Coles will have a total of 1,500 virtual hosts and 2,250 virtual machines in its 750 supermarkets.
Centralised Infrastructure Management: Coles worked with Xello to implement System Center products and technologies in order to centrally manage its physical and virtual server environments at its supermarket locations. One of the company’s key goals was to gain the ability to enable IT personnel to manage the geographically dispersed infrastructure from a central location. Xello played several important roles for this stage of the project. The company helped design the System Center environment that Coles used, helped implement the technologies, and also developed custom training materials to help IT staff at Coles manage the environment. “Our core focus at Xello is helping customers drive value from the System Center products,” says Neil Christie, General Manager at Xello. “System Center technologies are not just point products, but are best-of-breed solutions for systems management.”
At its data centers, Coles uses Microsoft System Center Configuration Manager 2007 to build and manage host servers at individual supermarket locations. Coles uses Microsoft System Center Virtual Machine Manager 2008 R2 to deploy, manage, and optimise its guest machines. It also uses System Center Data Protection Manager 2010 as a unified backup and recovery solution for each of the supermarket locations. “We use System Center Data Protection Manager to perform cross-server backups,” explains Morton. “That means that each of the two servers in the individual stores back up to each other and help ensure high availability of services.”
Coles also implemented Opalis, an automation platform for orchestrating and integrating IT tools. Opalis integrates and orchestrates the Microsoft System Center products and other management tools to automate best practices, such as incident response, change and compliance, and service–lifecycle management processes. Specifically, Coles uses Opalis in conjunction with Microsoft System Center Operations Manager 2007 R2. “We use System Center Operations Manager to proactively identify when there is a problem with our virtualized infrastructure and then use Opalis to automatically push that information into a proprietary logging tool,” explains Harmsworth.
Going forward, Coles plans to expand how it uses System Center solutions to manage its physical and virtual environment. For instance, in the long term, the company plans to use System Center Configuration Manager to also build guest machines for its host servers at the supermarkets. It will also evaluate Microsoft System Center 2012 Orchestrator 2012, the next release of Opalis. “We’ve barely scratched the surface of the capabilities that Opalis offers,” says Harmsworth. “We hope to explore more automation opportunities with System Center Orchestrator in the future.”