Do You Know Your Virtual Infrastructure "Inside and Out"?.
by Srinivas Ramanathan
July 24, 2009
Virtualization has been around for a while, and the business case for virtualizing IT resources has been clearly established; ranging from improving performance to enhancing scalability to better system utilization.
Why, then, does the idea that virtualization can serve an as entire enterprise infrastructure remain a new concept? This surprising result came to light from a recent survey conducted by Unisphere Research for SHARE, the world's largest association of corporate users of enterprise IT technology.
Among the survey findings was that most virtualization initiatives are scattered and few are enterprise in scope (Figure 1). About 22% of respondents say they are adopting virtualization on an enterprise scale, but 30% say it is selected on a department-by-department basis. Another 23% aren't even aware what their corporate virtualization strategy includes. Ultimately, however, a majority of respondents view enterprise virtualization as a long-term IT strategy.
Is Virtualization Enterprise-Wide, or Case-by-Case?"
Based on individual departmental initiatives
Have an enterprise-wide virtualization strategy
Considering an enterprise-wide strategy
Currently have no virtualization initiatives
Don't know/unsure
Other
30%
22%
18%
6%
23%
1%
Figure 1
Based on the survey results, one way to interpret the reason for the lack enterprise-wide virtualization adoption is the inherent complexity of managing business services and application performance at a level at least equivalent to that achieved in traditional physical infrastructures. Without that assurance - and consider how tricky it to manage physical IT resources! -- it is understandable why the majority of virtualization deployments remain siloed in individual departmental initiatives.
Knowing Your Virtual Infrastructure "Inside and Out"
Enterprise-wide virtualization deployment requires knowing how a business service is performing and which domains (network, server, VM, applications) are working and which are not. It is no longer sufficient to just monitor the uptime or resource usage levels of virtual machines (VMs) and physical servers and believe that the entire IT infrastructure is working well.
Most virtualization monitoring solutions just track the usage levels of physical server resources (CPU, memory, disk, network) and the relative usage levels of these resources by the VMs. This is an "outside" view of VM behavior. However, you also need an "inside view" to track such dynamics as end-user activity, resource allocation for each application and the application mix running inside the VM guest operating system.
Virtualized business services and applications have "no fixed address" like they did in the physical world. Whereas migration of a VM from one physical server to another may alleviate a problem for a short period of time, repeated migrations are often indicators of problems in the infrastructure. You need to be able to determine the root-cause of migrations before this escalates into a significant problem. If you cannot monitor VM behavior from both inside and the outside, you cannot take this diagnostic and corrective action fast enough. Hence, the "inside view" of a VM is critical for problem diagnosis and capacity planning - to know whether a high resource usage in a VM is being caused by excessive load, or by a malfunctioning application, or by inadequate capacity planning.
Because of skews between VM and physical machine clocks, it is well understood that VM clocks should not be relied upon for performance benchmarks. We are not advocating that this needs to change - the outside view of a VM is the true indicator of the performance of the virtual infrastructure. By using the "inside view" of a VM and comparing the relative resource usage levels among the applications running inside a VM, you can get a complete view of the virtual infrastructure and identify the root-cause of problems.
The eG VM Monitor - In-N-Out Monitoring for Virtual Infrastructures
The eG VM MonitorTM, from eG Innovations, is the only solution that offers such a holistic view. Using patent-pending In-N-Out MonitoringTM technology (Figure 2), eG VM Monitor uses a single monitoring agent to provide an outside view of how the VMs are using a physical server, and an inside view of what is happening inside a VM. This unique technology offers key insights into the cause of performance problems in virtual infrastructures from multiple suppliers, including VMware vSphere/ESX, Citrix XenServer, Microsoft VirtualServer, and Solaris Logical Domains.
The eG VM Monitor provides a comprehensive "outside view" of a physical servers, including the performance of the hypervisor, the service console and all of its virtual machines. Agent-based or agentless monitoring can be used; in the former case, eG agents need only be installed on the server -- not on individual guests. All the capabilities of agent-based monitoring are also available with the agentless monitoring option. The relative resource usage levels of the guest VMs show where the performance hogs exist.
To complement the outside view, the eG agent obtains an "inside view" that details the user activity, resource allocation and the application mix running inside the VM guest operating system. The eG VM Monitor automatically baselines all the metrics it collects, so that IT administrators can be informed proactively of any deviations from the norm. No other virtualization monitoring solution offers this combination of features.
Figure 2 - Monitoring VM guests: eG agents track the performance of each guest VM relative to
shared infrastructure resources (outside view) as well as the workload and application mix
of the individual guest VMs (inside view).
From a monitoring and management standpoint, the eG monitor for VMware infrastructures goes well beyond managing virtualized servers as discrete entities. End-to-end business service views show the applications and network devices that support each business service, and the inter-dependencies among them. Applications are associated with the virtual machines they run on, and each virtual machine is mapped to the physical machine upon which it is hosted.
The dependency of the virtual machines to physical machines is determined dynamically to support live migration technology. A patented root-cause diagnosis engine analyzes the service topology graphs and the virtual-to-physical machine mappings to pin-point where the problems areas in the infrastructure lie.
Armed with a tool such as this, the challenges of deploying virtualization on an enterprise-wide basis are far less daunting.
Srinivas Ramanathan is the founder and CEO of eG Innovations (www.eginnovations.com). eG Innovations is a global provider of performance monitoring and triage solutions for both virtual and physical IT infrastructures, and its eG VM Monitor won the "Best of VMworld" award in the application and infrastructure management category at VMworld 2008. Prior to eG, Ramanathan was a senior research scientist at Hewlett-Packard. He has extensive experience in Internet technologies, performance monitoring and management, and multimedia systems, having co-authored more than 40 technical papers and has been a co-inventor of 16 patents. He has a PhD in Computer Science and Engineering from the University of California, San Diego. Contact him at srinivas@eginnovations.com.