New IT technologies come and go on a daily basis, but sometimes they materialize into the real market and can become a mainstream disruptive technology. That’s what hypervisor technology did for server virtualization years ago. Today we see that desktop and server virtualization provide the flexibility to fulfill dynamic service delivery to the business. It’s no wonder that today’s IT managers tend to think of computing in terms of virtualized servers and infrastructures rather than physical infrastructures.
When it comes to management of such an environment, realistically, it takes a more comprehensive approach to ensure uptime and cost optimization of virtualized infrastructures. Back in the “physical” era, data center managers were challenged on physical capacity management around space, power and cooling. In today’s “virtual” era, physical capacity management is also combined with the demands of the virtual environment. It needs to be aligned with the capacity and distribution (location, clustering, load-balancing) of the physical environment.
Challenges with Virtualization Management Tools
Today’s data centers provide applications and services with several extra layers beyond what was present a few years ago:
- A services layer on top
- Applications on a virtual machine (VM) manager supported by hypervisor technology
- A physical layer of servers/storage underneath that to support it all
- A physical and agile (software defined) network
Managers that wish to run their businesses in line with their capabilities need to be able to manage the whole chain, from top to bottom and vice versa.
Current virtualization tools mainly manage the virtual environment. They know how VMs are provisioned and where they are, but they don’t take into account the physical servers on which the virtual environment is running. As a result, VM tools lack any contextual physical information.
Why is this an issue? Suppose 10 physical servers support part of the virtual environment. The virtualization management software can freely move VMs around on those servers. If the software puts a CPU load on a server in an area of the data center where the temperature is critical, the server could overheat and take down the VMs running on it.
Data center infrastructure management (DCIM) systems report temperature, power connectivity, location, and network connectivity – all the environmentals of the physical servers. They can assure the virtualization management software that the applications are hosted on a physical server runs at a healthy physical state. It can also forecast if it’s in the perfect physical condition to support new VMs.
There are three main reasons why managers want DCIM capabilities in a virtualized environment:
- They want to mitigate the risk of a server or network meltdown, so they need VMs on servers running in disparate healthy areas of the data center.
- They want to know the physical state of those servers: how they’re powered up, where they’re located and how they’re connected.
- They need to be as efficient as possible in managing both the physical, as well as virtual landscape of the data center.
The DCIM layer gives managers the visibility to only offload IT resources in those physical areas of the data center that can provide it, not only in terms of CPU capacity, but in environmental terms as well. After all, it’s better for the virtualization manager to launch new VMs in areas of the DC that are running cool, rather than hot. It’s also important to ensure adequate power and connectivity before trusting those servers to your VMs.
Here’s an example of how a DCIM can help IT manage a virtualized environment. A customer provides an IP address range for their data centers, and states that there are two virtual environments in this location with approximately 60 VMs supported by about 100 servers. An integration is performed between DCIM and VMware, and it is discovered that there were not two, but three virtual environments instances in the area. The integration shows the amount of energy usage and the list of active VMs. The company investigates and finds that the third hypervisor hadn’t been decommissioned, and it has 30 live and operational VMs on it. Had someone gone in and removed the servers supporting those VMs, the company would have lost some applications.
Integrating DCIM Systems
We see more and more that DCIM systems are getting integrated with VM management systems. Integration is typically accomplished by using the API of both the DCIM and the VM management systems to pull in data from the virtual environment, such as hardware, cluster, hypervisor and VM relations but also detailed information about CPU, storage, memory utilization and so on.
This allows managers to look at the hardware and also to spot trends or make predictions over time. For example, many VMs may be in operation at certain hours of the day and have certain capacity free at other times. From the hardware perspective, managers can track usage of the hardware irrespective of the VMs being managed on it.
Different DCIM vendors provide different levels of information about the physical plant. Some have bidirectional APIs where you can control the VM instance from the DCIM instance, or control the DCIM information from the virtualization management software. Other products are unidirectional and drive the virtualization management settings from the DCIM.
With today’s VMs running mission-critical applications, data center managers need to see the whole stack – from top down services to bottom up hardware – to ensure that the right servers are supporting the right VMs, applications and services. By integrating a DCIM with virtualization management software, data center managers can finally see the whole picture.