Virtual environments are subject to an extreme number of infrastructure configuration changes that increase the complexity in the IT enterprise. Knowing how network resources are connected, and the consequences of change, can make all the difference. Seemingly small changes to the virtual environments now bring potential risk to enterprise operations. Complex virtual environments coupled with high end-user expectations, the pressures to meet SLA agreements, and a focus on cost reduction have significantly increased the probability of disruption.

For example, virtual machines are typically connected to shared networks. Any virtual machine or group of virtual machines connected to a common network can communicate across those network links and can, therefore, still be the target of network attacks from other virtual machines on the network. As a result, administrators apply network best practices to harden the networks and isolate them to minimize the risks of data leaks from one virtual machine zone to the next across the network. The common approach used in implementing segmentation is the use of virtual local area networks (VLANs). This approach is based on VLANs providing almost all of the security benefits inherent in implementing physically separate networks. Two virtual machines in the network are unable to transmit packets back and forth unless they're part of the same VLAN. Proper segmentation of the virtual data center using VLANs requires careful planning of both the physical and the virtual network components. VLANs segmentation increases the network's complexity and creates environment dependencies between the physical and the virtual network.

Administrators require vigilant monitoring and tight coordination of configuration changes across both the virtual and physical network. A simple change to the VLAN configuration of the virtual infrastructure can impact the security level and the performance of VMs that will fail to perform the business task it was designed to do. Small variations of the initial configuration of the virtual network's VLANs structure or VLANs configuration changes to the physical network can produce large variations in the long-term behavior of the virtual and the physical environment. Just as Lorenz's Brazilian Butterfly has the potential to set off a tornado thousand of miles away, frequent changes that network VLAN configuration impose on the infrastructure can have serious consequences and negatively impact the security and performance of the entire data center.

Another example is VMotion migration technology, which adds a much needed dynamism to virtual environments. Virtual machines and their applications are moved between physical servers with features such as Distributed Resource Scheduling (DRS) and High Availability (HA) that are based on VMotion's enabling technology. DRS continuously optimizes resource allocations based on defined rules and resource utilization. The changes to resource allocation can be automatically executed by performing live migration of virtual machines through VMotion. The frequency of moves depends on the aggressiveness of the policy chosen. A more aggressive DRS implementation would set off a flurry of VM migrations on a frequent basis. Like Lorenz's Brazilian Butterfly, frequent changes that DRS imposed on the infrastructure can have serious consequences. If the butterfly hadn't flapped its wings and DRS stopped migrating VMs at such a high frequency, the trajectory of the system might have been vastly different. While neither the butterfly nor the DRS activities directly caused the tornado, the flap of its wings and the initial configuration of the virtual environment are an essential part of the initial conditions that presage disaster. Recurrence - the approximate return of a system toward its initial conditions - together with sensitive dependence on initial conditions are the two primary ingredients for chaotic motion.

The Butterfly Effect is a real phenomenon in virtual data centers that can increase risk, drive up costs, negatively affect availability, and fracture IT's credibility. Several core problems faced by nearly every large IT organization have magnified these burdens. Information about virtual configuration interdependencies is poorly understood, tracking it is inefficient, and mapping out a dynamic virtual infrastructure is a complicated and time-consuming endeavor. But organizations can mitigate these effects by automating their virtual infrastructure management.

IT organizations need a new breed of management tools that understands and takes advantage of virtualization's unique requirements and capabilities. Virtual Infrastructure Management is now being embraced by organizations around the world as a way of achieving large increases in availability and reliability for relatively modest investments in services and tools. Organizations implementing tools like Reflex' management solution for virtual environment have achieved these goals by reducing the time required to resolve problems, improving capacity planning, monitoring configuration changes, increasing business resiliency, and better managing their virtual data center assets. Those enterprises that have made the leap to effective Virtual Infrastructure Management no longer fear the winds of change or Lorenz's Butterfly Effect. Leveraging the Virtual Infrastructure Management tools, IT departments can evolve and add the following capabilities:

• Gain visibility and assert control across virtual environments. Since you can't control what you can't see, visibility is key in managing and monitoring the virtual infrastructure. Administrators need a visual representation of their virtual environment to understand the entire virtual network, track changes, and address virtualization challenges that have management implications.
• Track all infrastructure changes in real-time and maintain an accurate model of the virtual environment. With Virtual Infrastructure Management tools at their disposal, administrators can track changes to assets and alert stakeholders when critical assets are modified. In addition, the administrator has access to historical changes to identify unauthorized configuration changes, configuration errors, and enforce policies across the entire virtual infrastructure. Each time a change is detected, the change and all relevant details (who, when, why) are recorded
• Monitoring and managing virtual performance enables organizations to increase business efficiency and assure business continuity in the virtual data center. The ability to detect network bottlenecks inside the virtual infrastructure, identify over-utilized VMs, improve the performance of critical applications, and avoid network outages will allow administrators to improve the delivery of these critical applications, quickly troubleshoot issues, and optimize their virtual infrastructure.

From a philosophical standpoint, the Butterfly Effect provides a framework for understanding how seemingly small and inconsequential events can have radical consequences down the line. This theory can be applied to the virtual data center, a system that has become exponentially more complex and unpredictable as it grows in scope and gets beyond our direct control. While such systems will always possess a measure of uncertainty, we will require new tools and diagnostics that let us look deep into the black box.