Frequently Asked Questions (FAQ)
The virtualization management space is becoming somewhat crowded lately, with a number of companies either migrating their management solution over to support virtualization or new companies starting up to try and answer the management issue. How does the eG VMware Monitor differ from other products in the virtualization space?
Many solutions in the virtualization space focus on monitoring the virtual environment as a “silo”. These solutions extract metrics using instrumentation/APIs available in VMware ESX and vCenter, and even though these metrics may be incorporated into a common console where metrics for physical servers, networks, and applications in the infrastructure are also available, there is little or no correlation between the virtual infrastructure’s performance and the performance of the applications that rely on the virtual infrastructure. The eG Enterprise also relies on the in-built VMware instrumentation for insights into the VMware infrastructure. However, it differs from other VM monitoring solutions in several key ways:
- Firstly, the ESX infrastructure has only an “outside” view of each of the VMs supported by it. The VM Kernel does not know the details of what is going on within a VM - e.g., if the CPU usage of a VM is high, is that because the workload is high, or because there is a single run-away process that is taking up all the CPU resources? To effectively monitor a virtual infrastructure, it is essential for a monitoring system to have an “inside” view of each VM - to know what is happening within the VM. eG Enterprise includes a patented In-N-Out monitoring technology wherein a single agent is able to monitor the ESX server, the service console, and provide both an inside and an outside view of each of the VMs.
- Each of the metrics collected by the eG agents can be automatically baselined - past history is used to determine time-varying norms for the metrics, and alerts triggered when a bound is violated. By avoiding the need to set manual thresholds for thousands of metrics, auto-baselining as supported by eG Enterprise can save hundreds of man hours of consulting time. This capability is also critical for an IT infrastructure to be managed proactively - so alerts are generated before users complain.
- Most monitoring solutions monitor virtual environments as silos. In reality, the virtual environment is an integral part of an end-to-end business service. For example, a banking application may include a web server front-end, a middleware application server, and a database backend. The database server may be running on a VM that is hosted on an ESX server. If another VM on the same ESX server has an I/O intensive application (e.g., a media server) running on it, the I/O intensive application may impact the performance of the database server. Inter-dependencies in a virtual environment (which applications are running on VMs that are co-hosted on the same ESX server) play an important role in the performance of the end-to-end business service. eG Enterprise analyzes the performance of the virtual infrastructure in the context of the business services that it supports. Inter-application dependencies as well as inter-VM dependencies are taken into account when eG Enterprise diagnoses where the root-cause of a problem with a business service lies.
Usually you hear things like “just let VMware’s DRS, HA and VMotion handle the resources and move the virtual machines around for you to keep them up and operational.” Why then do you find it important to track VMotion server moves?
VMware DRS and VMotion are very good technologies that enable an adaptive, dynamic environment. However, the dynamic nature of the virtual environment makes proper monitoring and management a challenge. For example, a database server may be on ESX server A at one instant of time and on ESX server B at another instant of time. Hence, dependency mappings used for root-cause diagnosis (as discussed in the first question) are no longer static. The monitoring system needs to be able to dynamically determine which VM is running on which ESX server.
VMotion/DRS are based on pre-specified criteria - if these criteria are set too aggressively, this may result in unnecessary movement of virtual machines across ESX servers. Such situations need to be monitored and remedied.
One of the key questions that comes up with VMotion/DRS is - why did a VM move? Since it tracks what is happening at every layer of the virtual infrastructure, eG Enterprise has the information that administrators can use to answer this question - I.e., what happened on the ESX server prior to the VM movement - what triggered this?
What are some of the challenges with monitoring VMware infrastructures?
VMware technologies and virtualization in general aim to make better use of existing hardware resources, by running multiple VMs on the same hardware. Any time there is resource sharing, there are also issues of resource contention. In a VMware environment, the performance of a VM may be impacted by another VM running on the same physical server. These dependencies have to be taken into account when monitoring and managing this infrastructure. The use of VMotion and DRS makes the problem even harder - since VMs could be moving from one ESX server to another.
One of the benefits of virtualization is the speed with which new VMs can be provisioned to address the business imperatives of flexibility and rapid response to changing service requirements. This of course also carries the widely discussed risk of server sprawl, with its potential for loss of control and oversight. In these times of strict corporate governance and regulatory compliance, the monitoring solution needs to give management absolute confidence that they are in control and are monitoring the overall infrastructure in an appropriate manner.
Can you tell us more about the information you are gathering from the virtual machines themselves? Are agents involved or not?
Knowing what is happening within the VM is key to effective root-cause diagnosis and performance management in a virtual infrastructure. Recognizing this, eG Enterprise has incorporated a patented In-N-Out monitoring technology that includes in-depth insights into what is happening with the VM guest, simply by installing a single agent on an ESX server. Without requiring an agent on every guest VM,
eG Enterprise is able to determine what processes are running inside a VM, which are the top resource consuming processes, which user is logged on to the guest VM, etc.
Is your solution specific to VMware only or does it support other applications?
eG Enterprise offers monitoring of the virtual as well as the physical environment. Support for over 150+ applications (Citrix, terminal services, SAP, Active Directory, SQL, Oracle, Sun ONE, and many others) and 10+ operating systems (Windows, Linux, Solaris, HPUX, AIX, etc.) is available in the eG Enterprise - see http://www.eginnovations.com/web/platforms.htm for details.
Are you supporting vSphere/ESXi?
Yes; eG Enterprise provides monitoring support to vSphere/ESXi. Whereas ESX 3 and 3.5 have the service console where an agent can be installed, vSphere/ESXi does not have the service console. Hence, eG Enterprise provides a completely agentless monitoring solution for vSphere/ESXi. This solution has almost all the capabilities of the local agent in ESX 3 / 3.5.
How accurate would your “inside view” be, considering the fact that the software clock that the VMs support is different from that of the physical ESX server?
VMs are known to suffer from clock skews and hence, metrics collected from within a VM have to be analyzed with care. Rather than use metrics collected from within a VM in absolute terms, eG Enterprise uses the inside view to compare the resource usage levels across applications running in the VMs. Other key indicators of VM and application bottlenecks such as memory leaks, handle leaks, disk partitions filling up, etc., can also be obtained using the inside view of a VM.
Do you plan to support other virtualization technologies?
Yes - our goal is to be the leading provider of virtualization management software - and towards this end, we are already supporting Citrix Xen, Microsoft Hyper-V, Solaris LDom, Solaris Containers, and AIX LPAR technology as well.
How is the product priced?
eG Enterprise is priced per VMware ESX server being monitored - i.e., the licensing is not per CPU, per socket, or per VM. This provides administrators insights into the performance of the VM kernel, the service console (if applicable), and the individual VMs and the applicationsthat run on these VMs.
Does the product have a web-based interface?
Yes - eG Enterprise follows a 100% web-based architecture - the agents communicate with the management console using the web protocols (HTTP/HTTPS) and the only way to access the management console to view alerts, pop reports, etc. is using the web browser.