Virtualization has enjoyed mass adoption -- half of all server workloads run on virtual machines (VMs) today -- thanks to efficiency gains and cost reductions it can achieve through server consolidation. The rise of cloud services will push virtualization closer to ubiquity, as IDC predicts 70% of all workloads will run on VMs by 2014. This explosive growth, however, carries a potential risk for cloud providers: VM sprawl.
The consequences of VM sprawl may negate the key benefits that entice service providers to adopt virtualization. An ever-growing pool of VMs can reverse any efficiency gains once systems management becomes unwieldy. Additionally, cost savings will vanish if too many VMs demand excessive resources from their hosts. An environment with gaps in IT management inherently exposes cloud providers to security threats as well.
Controlling VM sprawl in a cloud starts with awareness. Only when a cloud provider understands how to recognize VM sprawl can it be curtailed. Unfortunately, there are many misconceptions about VM sprawl in the cloud that could interfere with effective management. To clear up some of the more common myths about VM sprawl, we examined the fact and fiction surrounding VM sprawl in the cloud.
Four myths about VM sprawl in the cloud
Myth: Virtualization inherently facilitates much more efficient resource utilization than physical servers and appliances can achieve, so VM sprawl in a cloud environment really is a non-issue.
Reality: Part of what draws customers to cloud services is the flexibility of on-demand access to IT resources to meet fluctuating capacity requirements. Cloud services enable customers to stop over-provisioning on-premises servers, but transferring server sprawl to the cloud provider's environment is still a risk. After all, cloud providers can deploy new virtual machines in a fraction of the time and -- thanks to cloud automation -- with fraction of the effort it takes to deploy a physical server. Consequently, customers may tend to be more liberal in provisioning new VMs. This click-and-configure capability makes it almost too easy for cloud providers to spin up new VMs -- unlike physical server deployment, which requires unpacking the new machine, racking and stacking it, and then cabling it and configuring it.
Cloud providers believe they can overlook spikes in new VM deployments because capacity requirements ebb and flow in the cloud. Unfortunately, allowing a rapid proliferation of new VMs can create a cloud environment that is anything but optimal. Efficient as these virtual machines may be, they still require IT support, effective security and compliance management, storage and other resources. Cloud providers can quickly find they are supporting an environment populated by VMs that are underutilized and cost more in licensing and administrative support than they can justify.
Resolving the issue: Cloud providers need good governance to prevent VM sprawl and a pragmatic way to execute on these policies. Practices that enforce things like consistent VM configurations and continued VM monitoring help ensure that cloud providers are able to not just monitor all of their resources but also to manage them throughout the entire lifecycle.
Myth: VM sprawl in the cloud is inevitable because separate groups within the same enterprise put redundant data into on-demand environments.
Reality: Because cloud purchasing decisions are often made on the departmental level, there is a real possibility that multiple groups within one organization may put the same workloads into different clouds. Customers that rely on multiple redundant clouds may soon lose track of the data and use a provider's cloud environment inefficiently.
Resolving the issue: Cloud providers have a key advisory role to play in helping enterprise customers take a more strategic and holistic view of their cloud strategy, and policy is crucial here.
Although individual departments within an organization have some degree of autonomy in their cloud deployment decisions, providers should ensure each deployment complies with the customer's corporate policies dictating where specific applications can be hosted. Cloud providers should also communicate to other groups within the organization when a workload has been deployed in the cloud. Cloud providers can help customers define, implement and execute cloud deployment policies by counseling clients as well as implementing monitoring tools and techniques to ensure compliance with these mandates.
Myth: VM sprawl in the cloud can be measured by just looking at number of VMs.
Reality: Although numbers do matter, there is no universal calculation that can be applied to quantify VM sprawl in the cloud. This makes curbing VM sprawl very difficult. Some providers have tried to shut down VMs en masse to determine which machines are redundant or underutilized -- based on end user reaction -- and have then decommissioned those machines that aren't deemed essential.
Taking the redundant machines offline doesn't solve the problem of VM sprawl in the cloud, however. These machines still incur licensing fees, but they no longer receive automated security patches, which causes them to miss crucial updates and expose the customer to vulnerabilities.
Resolving the issue: Cloud providers should work with enterprises to limit the amount of time a virtual machine remains in limbo, setting a policy that eliminates an offline machine entirely from the network after a standard period of time.
Myth: Managing virtual machines is just so much more challenging than containing physical servers that it is impossible to contain VM sprawl in the cloud.
Reality: While virtualization consolidates the number of physical devices on the network, it also usually increases the number of applications and operating systems, thus making the environment more complex to manage. But complex doesn't mean impossible.
Although it may be easier in many ways to rein in physical server sprawl because of the physical limits of the facilities themselves, this doesn't mean virtual environments are destined for rampant, uncontrolled propagation. With the right set of tools -- and a cloud provider with expertise in managing combined virtual and physical environments -- VM sprawl can be controlled.
Resolving the issue: Conventional management tools and practices designed to support physical assets -- such as trying to track them manually using spreadsheets -- are typically insufficient in a virtual environment. Cloud providers should evaluate automated information lifecycle management tools that enable them to keep track of new deployments. Providers should also advise customers to limit the number of users who can provision new VMs. There should also be a standard VM image used to maintain the integrity of the environment.
About the author: Amy Larsen DeCarlo is a principal analyst at Current Analysis, where her research focuses on assessing managed and cloud-based data center and security services.
Take the quiz: Now that you are done reading this tip, take the quiz (Quiz: How can you control VM sprawl in the cloud?) and find out if you understand the VM solutions needed to keep VM sprawl at bay.