The reality is that none of your employees need to lose their jobs. You can call a water cooler meeting and reassure them all that their positions are secure as long as they can adapt to a few changes. Let's examine three areas in which virtualization will change an administrator's job: building servers, patching and monitoring.
Building servers Service creation
First of all, the phrase must change from Building servers to Service creation. This change is necessary because virtualization completes the circle of commoditization. I am a fan of the philosophy of Tim O'Reilly that software should be a commodity upon which services are built. Virtualization takes the physical host out of the picture so that all that is left is the service. Administrators will need to stop thinking in terms of building servers, but rather creating services. This does not mean that someone whose job was to build servers is now out of work. Instead, it means that person's job is now to create services as they are needed.
For example, imagine that a constituency at work wanted to start an intra-office blog. Prior to virtualization, the first step would have been to task someone to build a dedicated Web server to host the blog. This could take up to several hours, depending on the readiness of hardware, OS installation and configuration, and application installation and configuration. With the advent of virtualization, the same task simply involves deploying a pre-built virtual machine (VM) image. This image already will have been configured as a Web server and perhaps even with the blogging software installed and configured as well. What is left is the need to further customize the blog to the customers', in this case coworkers', wants and needs. The administrator whose job was to build servers must now transform his skill set into being able to create and customize services, a much more efficient and productive use of his time. If this task is already assigned to someone in this administrator's group, fear not, there is still patching that needs to be accomplished!
There is a misconception that the advent of virtualization brings with it the end of patching. This is simply not true. Virtualization can make the regression testing of patches easier by enabling rollback, but it certainly does not halt the necessity to apply patches; indeed, it creates some additional work. Typically a system administrator waits for patches to be released, evaluates them and then applies those she deems necessary to her servers. This routine is slightly altered when virtualization enters the mix. Instead of applying patches in one place, the administrator must be apply them in three places: the host server, the VM and the VM template (VMT) and/or golden master (GM).
All virtual machines run on a host server that provides a virtual machine monitor (VMM), and that host server at some point will require patches. VMware provides all of the patches necessary forVMware ESX Server 2.5.3. With regards to XenEnterprise, according to Simon Crosby, CTO of XenSource, "XenSource will fully support the whole product (including the Dom0 OS)." If the VMM requires a separate host OS such as Microsoft Virtual Server 2005 R2, VMware Server, or a separate control OS (dom0) like Xen 3.0, then the system administrator will need to patch the host or control OS as well.
The system administrator must also patch the VMs. She is familiar with this process because it is the same as patching the servers that are in existence today.
Finally, the VMTs and/or GMs will need to be kept up to date on patches as well. A VMT is a configured VM that has been turned into a deployable template with software such as VMware's VirtualCenter or Xen-tools (Xen-tools refers to its templates as images, but they have the same functionality as templates, so I will refer to them as templates). Likewise, a template could be generated by configuring a VM and then saving a copy of it to another location, and then using that copy as a starting point for future like VMs. This is called a GM. VMTs and/or GMs are kept up-to-date on patches by patching deployed versions of themselves and then recreating the VMT and/or GM. Xen-tools even has a nifty feature that lets you apply software updates directly to a template.
The previous two examples of service creation and patching should seem familiar to most IT managers. The changes I outlined are probably very similar to the adjustments that had to be made when software such as Symantec Ghost was first introduced. But monitoring a virtualized environment is brand new and more important than ever before. IT professionals who work in an environment that employs mainframe computers are used to the eggs-in-one-basket situation, in which a single hardware of software fault can result in all systems going offline. A single virtual server may play host to tens of VMs, meaning that if the host hardware of host VMM software has problems, the effect is much greater than if a single Web server were to go offline. This is why it is paramount that an administrator of virtual services know how to monitor his environment. Monitoring a virtualized infrastructure can be broken down into three parts: monitoring hardware, monitoring the VMs and monitoring the VMM.
Monitoring a physical server is a task we all do now, and monitoring a virtual Web server requires the same tool set and skill set as monitoring a physical Web server. The new kid on the block is monitoring the VMM. The VMM could slip into a degraded state without showing up on the hardware or VM monitors until it is too late. An administrator will need to learn how to monitor the VMM and what indicates trouble.
I hope that I have shown you that just because virtualization can make creating services and regression testing of patches a breeze, it does not mean that there is less work to do. Your employees may not have to work harder, but they will have to work smarter.
Dig Deeper on Server virtualization staffing, careers and budget
Related Q&A from Andrew Kutz
A user wonders how well Ubuntu will serve him/her in terms of stability, and gets release recommendations from an expert. Continue Reading
Learn about an emerging product that aims to decrease time spent fixing dependencies. Continue Reading
Site expert Andrew Kutz answers a user question on creating shared storage in a VMware ESX Server environment. Continue Reading