Christie, with headquarters in Cypress,California as well as in Kitchener,Ontario, manufactures a variety of display technologies and solutions for cinema, large-audience environments, control rooms, business presentations, training facilities, 3D and virtual reality, simulation, education, media and government. Virtual Servers Popping Up Everywhere The IT environment supporting Christie operations includes datacentres in both Cypress and Kitchener running mission-critical applications such as the JD Edwards ERP, Microsoft Exchange, Microsoft SharePoint, Agile Enterprise Content Management and an Amdocs application used by the Managed Services division to monitor and manage service delivery for customer projection systems. A significant challenge the Christie IT team faced related to this infrastructure was a new twist on an old theme – server sprawl – but in this case, virtual server sprawl. Beginning with a single, small VMware ESX host server deployment, the company had made a substantial investment in its virtualization journey, which progressed quickly to include 20 physical server hosts, with another half-dozen planned. Although virtualization provided a much higher level of agility to respond to the company’s needs for computing services, one thing that always seems to thwart their virtualization efforts was hosts running out of memory, which is why they needed so many hosts. “As soon as I deploy a new host server, virtual machines start appearing quickly and the host soon runs out of memory – it’s like trying to build a railway just ahead of a train barreling down the track,” laments Scott Elliott, Supervisor, Network & Systems at Christie. He explains this virtual server sprawl by suggesting that, “Virtual machines are almost too easy to provision and carry the perception of being ‘free’, so users want lots of them.” Scaling Upwards, Not Outwards To handle virtual server growth and to solve the problem of continually running out memory on host servers, Christie accepted a recommendation by Compugen to increase server capacity through the addition of four large IBM System x3850 X5 servers at each of its two datacentres. The IBM x3850 X5s, which are based on high-performance Intel Xeon processors, are 4U enterprise servers expandable up to eight CPU sockets and 80 cores, and have up to 2TB of memory, further expandable to 3TB through use of memory expansion modules. “We started our virtualization journey using IBM System x3650 servers, which served us quite well initially, but our growth and evolving requirements necessitated that we move onto a different technology,” states Elliott. Rather than expanding outward by adding more x3650s, they decided to contract inwards, by consolidating onto fewer but larger x3850 X5 servers. The x3850 X5s provide the CPU horsepower, and more importantly the memory required, to support a much larger server consolidation ratio, thus supporting more virtual servers than the previous server infrastructure. Elliott admits to having previously been reluctant to deploy applications that require multiple CPUs because of multi-threading or attempt to virtualize some of the bigger workloads such as the JD Edwards application. “I’ve been burned by CPU contention in the past, but with the new IBM servers, I don’t have to worry about these large applications draining all our CPU resources or the memory needed for virtualization,” claims Elliott, citing as an example the fact that their Web front end alone for JD Edwards involves seven virtual servers. TheIBMservers are now running Christie’s tier-one virtualized production workloads, including JD Edwards, Agile ECM, SharePoint 2010, Exchange 2010 and eventually the all-important Amdocs service management application as well. With the help of VMware’s vSphere and vCenter Site Recovery Manager, the new IBM servers in both datacentre locations are also providing a robust disaster recovery platform for Christie. Consolidating many smaller physical servers into just four large boxes has also resulted in a smaller physical footprint, requiring less system administration and server management, and a larger logical footprint, making it easier for the IT team to tackle new projects and deploy new applications.