This is the first in a three-part series that examines the challenges for administrators as they scale and manage their virtual infrastructures. This first tip outlines the typical phases of a virtual infrastructure deployment and details the first challenge: Virtual server sprawl.
In their early stages, most server consolidation projects yield favorable results, but customers find that server consolidation benefits don't always translate into larger installations. As users expand virtualization deployments, they often encounter obstacles in terms of scaling and manually managing virtual infrastructure. In this discussion, we explain how this central virtualization management challenge develops, how manual methods to track and manage virtual machines (VMs) become problematic in a fast-growing deployment, and some virtualization management tools that can minimize the problems of sprawl.
Phases of virtual infrastructure implementation
Organizations that invest in virtualization move through four phases of deployment that are characterized by certain qualities and problems. During these phases, the value to an organization grows, but so do virtualization's management challenges (see Table 1).
Table 1: The four phases of virtual infrastructure deployment
Server sprawl and management issues
Despite IT's efforts to anticipate and control virtual server sprawl, the ease of cloning virtual machines (VMs) can cause them to multiply. Many organizations have discovered that when a server becomes virtual, it becomes more difficult to identify and track. Based on the Taneja Group's discussions with vendors and end users, more than 50% of companies suffer from serious bouts of sprawl during the first year of server consolidation initiatives.
Unlike the flu, server sprawl cannot be shaken off easily. Following an initial wave of server consolidation, enthusiasm often sets in as organizations discover how easily they can deploy new VMs. But even when only a small number of people can create VMs, virtual servers can proliferate throughout an organization if left unchecked.
Sprawl has a serious side effect: As virtual machines propagate, organizations lack governance policies to configure these VMs. In the absence of standards, users and administrators configure virtual machines on spec, often overallocating resources to VMs, which creates waste and inefficiency. This problem is exacerbated in a virtual desktop infrastructure (VDI), in which large volumes of VMs are often created over a short period of time.
Managing VMs with manual processes
How do most IT organizations manage VM sprawl? Not surprisingly, they tend to turn to tools and processes with which they're familiar. Most administrators, for example, use traditional spreadsheets to track and manage virtual servers. Administrators log information such as VM owner, resources allocated, approved usage and lease period to track an organization's total number of VMs and their workloads.
While a traditional spreadsheet is a temporary management solution, most administrators can't keep up with VM additions and changes in a growing virtualization deployment. What happens when a developer creates a new virtual machine for testing purposes and then moves it to a test environment? What happens when the same developer moves the test VM back to the general population?
When virtual machine migration tools like VMware VMotion enter the mix, manual tracking can become even more difficult. One IT administrator reported that although he spent a vast amount of his time tracking VMs, his Excel spreadsheets were woefully out of date. As an organization expands its virtual infrastructure, a manual approach to tracking and managing VMs suffers from error and inefficiency. Even when spreadsheet data is current, it's usually only accessible to one or two administrators.
Even faithfully executed manual processes such as these are doomed to fail in a rapidly growing virtualized environment. As a result, management and operational costs soar because of highly inefficient administrative processes and the suboptimal allocation of system resources. Embotics Corp., a virtualization lifecycle management vendor, has modeled and measured the typical costs of virtual server sprawl. Based on its findings from end users, the costs of sprawl can offset an organization's savings from server consolidation.
Fortunately, new products are poised to enter the market to help virtual server administrators better cope with the growing pains of virtual infrastructure deployments. Marketed as virtual infrastructure management, lifecycle management, policy-based management or orchestration applications, these products go beyond the capabilities of hypervisor element managers to solve many of the problems associated with fast-growing virtualization environments.
Controlling server sprawl with VM lifecycle management
Virtual infrastructure management products help to address virtual server sprawl by enabling users to produce well-defined processes for provisioning new VMs. With these products, only authorized administrators can create and provision new virtual servers. In the case of DynamicOps Virtual Resource Management (VRM), users with similar needs and responsibilities are placed in the same provisioning group that dictates the amount of resources and types of VM-building templates that are available to new virtual servers. VM templates include workflow policies, security profiles and specific build configurations.
These build templates eliminate the errors of a purely manual approach. A global financial services company using this product found that the built-in process and VM templates helped reduce the number of new VMs being created to those required for specific business purposes.
DynamicOps VRM also assigns expiration dates to new virtual servers, ensuring that they are automatically decommissioned when they reach the end of life. With this feature, the same end user successfully retired extraneous VMs, particularly in the development and testing functions, on an ongoing basis, thereby reducing the effect of sprawl. As VMs are retired, the system resources originally allocated to them are reclaimed and repurposed, improving overall resource efficiency.
In part two of this series on virtualization management challenges, we explore other obstacles as users scale a virtual infrastructure and discuss how new technologies have helped them combat these issues.
ABOUT THE AUTHOR: Jeff Byrne is a senior analyst at the Taneja Group, an analysis and consulting group focused on storage and storage-centric server technologies with a concentration in the developing area of virtualization. Byrne can be reached at firstname.lastname@example.org.
And check out our Server Virtualization blog.