British fashion designer Paul Smith has a global network of retail outlets managed by 15 IT staff members in two data centers in the United Kingdom. It wanted to expand its virtualized environment and embrace cloud computing to further increase efficiencies and contain costs. It used Windows Server 2012 and Hyper-V to virtualize business-critical applications—and achieved a 10 percent improvement in their performance. Data center administrators can run almost triple the amount of virtual machines per server and provision new servers in a day, instead of four weeks. Paul Smith expects to virtualize 30 percent more of its servers by the end of 2012 and to save more than £840,000 a year in IT costs. It can also transfer one full-time equivalent from data center administration to work on IT projects that drive innovation in the business.
Situation Nottingham-based designer Paul Smith showed his first menswear collection in Paris under the Paul Smith label in 1976. Today, there are 14 different collections—produced in England and Italy—under the Paul Smith brand. In addition to the company’s 17 shops in England, Paul Smith shops are found in fashion capitals around the world, including 200 locations in Japan. Each one is a showcase for diverse objects that complement the clothing collections with a selection of jewelry, books, art, and antiques.
The Paul Smith IT department consists of 15 people who support the business on a global scale from the Nottingham head office. The centralized IT model works on a three-tier structure, with head office and the company’s UK-based data centers as Tier 1; branch offices in fashion capitals such as Paris, Milan, New York, and Tokyo as Tier 2; and the company’s 35 retail stores as Tier 3.
“Our core business has more than 1,200 employees and from an IT perspective our team is relatively small for the services we provide to the business,” says Lee Bingham, Head of IT at Paul Smith. “We have two primary data centers in Nottingham and a satellite data center in London.”
From its inception, Paul Smith has worked with Risual, a member of the Microsoft Partner Network with Gold competencies, to build a cost-effective, interoperable IT infrastructure that delivers real business value. To that end, the company has standardized on Microsoft technologies. “Aligning our business with Microsoft and working with Risual has been an extremely successful strategy,” says Bingham. “While we have made great progress with optimizing our data centers, and the level of service that IT delivers to the business is already high, we are always interested in the latest technologies to improve efficiency, lower costs, and better serve the business.”
Virtualization Goals In April 2008, to provide highly available applications and services to the business and to boost disaster recovery capabilities, Paul Smith deployed a geographically redundant virtualized environment in its primary and secondary data centers. The IT teams from Risual and Paul Smith used the Windows Server 2008 R2 operating system and Hyper-V virtualization technology, configuring eight HP ProLiant DL 580 servers in a high-availability cluster replicated at both sites and hosting 225 virtual machines, for an average of 28 virtual machines per host. At Tier 2 locations, the company used Hyper-V running on a single server to host on average 15 virtual machines to satisfy local IT requirements. To replicate business data back to the data centers in the United Kingdom, the company uses Distributed File System (DFS) Replication, a component of the Windows Server 2008 R2 operating system. DFS Replication is a multiple-master replication engine that administrators can use to keep folders synchronized between servers across limited bandwidth network connections.
By 2011, Paul Smith had virtualized approximately 50 percent of its servers. “We had an HP EVA 6100 storage architecture that allowed us to failover virtual machine workloads between our geographically dispersed clusters within a five-second window, and we used Microsoft System Center data center solutions to administer both physical and virtual servers,” says Bingham. “We had come a long way, but we wanted to drive down costs even more—the physical servers, the power, the hardware maintenance—by virtualizing up to 80 percent of all our servers.”
However, Paul Smith was unable to virtualize applications that were based on Microsoft SQL Server data management software and certain important line-of-business applications, such as Cognos, the company’s business intelligence solution, because of their criticality and performance requirements. “We had to keep some applications running in a physical environment because Hyper-V in Windows Server 2008 R2 supports virtual machines with a maximum of four virtual processors and 64 gigabytes of memory,” says Bingham. “We needed a more scalable virtualization solution so we could reach that 80 percent benchmark of virtualized servers in our data centers.”
Achieving that goal would help Paul Smith IT staff respond more quickly to business needs and save the company money required to procure and manually build physical servers. ”With physical workloads, additional capacity requires additional hardware,” says Bingham. “By the time we source the hardware, wait for delivery, and provision a server for service it could take four weeks. And every new machine costs approximately £5,000. We also pay for administration labor. We wanted to minimize these costs by virtualizing more servers, even those hosting applications that run on 16 or more processors.”
Paul Smith also wanted the performance and reliability of technologies such as network interface card (NIC) teaming. This is the process of combining two or more physical NICs into one single logical NIC, which can be used for network fault tolerance and load balancing. “Paul Smith used a mixture of NIC vendors and there were time-consuming compatibility issues that forced us to segment network configuration among cluster hosts,” says Richard Proud, Director at Risual. “Instead, Paul Smith wanted to converge networks for better bandwidth and manageability.”
Plans for Private Cloud Once it moves its non-virtualized workloads into a virtual environment, Paul Smith plans to exploit the additional agility and automation benefits that it would get by building private clouds. IT staff can enable a more responsive, on-demand allocation of pooled IT resources—such as applications, networks, servers, storage, and services—for employees to use when and where they are needed.
“We see private cloud as the next step in our virtualization journey,” says Bingham. “Building a private cloud infrastructure will enable us to be more agile in responding to business and end-user requirements, through self-service capabilities and automation. We wanted the best technology in place to make that a reality.”
To learn more about this private cloud solution Paul Smith used have a trial of:
System Center 2012
Windows Server 2012
Read the rest of the Case Study to see how Paul Smith worked out a strategy to transform the data centers into a cloud platform capable of supporting large, high-performance virtual machines.