At a VMUG meeting in Minneapolis in December, VMware employees mentioned that Virtual Infrastructure would be getting a new name, vSphere. A few days ago, VMware officially announced vSphere, the successor to the Virtual Infrastructure (VI) product line. VMware is hailing vSphere as the first true cloud-based operating system, hoping to capitalize on the hype that surrounds cloud computing.
VMware has had products available for years that allow administrators to cluster resources and place virtual machines on a virtualized abstraction layer that spans multiple hosts, pooling RAM, CPU and other system resources. When we had heard there was a raging debate about whether a private cloud was possible, we immediately though of all of our successful implementations of the VI product. vSphere is designed from the ground up to sit on low cost and energy efficient computing resources and allow for the flexible deployment of systems onto the cluster. This allows organizations ranging from small businesses to enterprise, from education to government to deploy new data protection and high availability resources, to pool IT assets in a manner not previously available.
The key components of vSphere all not all new. ESX and ESXi are the hypervisor. These sit on the physical machines (aka the Hosts) and build the virtualization layer. Sitting on top of the hypervisors is vCenter Server, which allows for the actual provisioning, monitoring, physical to virtual conversion process and centralized management. The vCenter Update Manager keeps all of the ESX systems updated (as well as some of the VMs themselves to help reduce the surface space of update management). The VMware High Availability piece gives failover between hosts. VMsafe is a another component that provides security APIs; while offerings from 3rd party developers are fairly immature expect this to grow rapidly as the virtualization industry moves into its next stage.
vSphere was built for microprocessers. The Nehalem and its successor, Westmere, are designed with collaboration from VMware; as such, they are built for virtualization. When you are looking to plan for a potential upgrade to vSphere, it’s important to keep in mind that each member of a vSphere cloud is going to run at the speed of the slowest host. Therefore, you will have tiers of VMware virtualized clouds, each with a class of system in it (for larger environments). The Nehalem and Westmere are designed for 8GB of RAM, so you’ll want to make sure to put plenty of memory into the cluster nodes, which have a deminishing return on investment (in terms of memory) around 120GB (so don’t be afraid of going hog wild on the memory front, those VMs need it!).
Overall, our tests of vSphere have shown a considerable performance gain for the guest operating systems running on hosts with newer hardware. Older assets have a lower impact on performance, but still have a slight upgrade. The biggest management features that we’re finding useful are an upgraded vCenter (for converting those physical systems over to virtual hosts), enhancements to Vmotion and automation. With the latest tools it is fairly straight forward to automate nearly every task using vCenter, including the deployment of new virtual machines based on templates, restarting a virtual machine and migrating them using Vmotion.
While the vSphere product may seem overwhelming at first, it begins to bring into focus a contained and mature VMware based infrastructure. There are a lot of new features; but there is bound to be a lot of marketing spin and while I’m sure it can, out of the box vSphere will not do your laundry. In order to help guide you through the planning phases of the next generation of the data center (which is after all, the true target of vSphere 4), 318 is here to provide the experience you need with regards to VMware licensing, architecture and of course support – be it with the guests, the hosts, the storage layer or the virtualization layer itself!
