VARs Look For Power And Profit With Virtual Machines5:39 PM EST Tue. Aug. 30, 2011
Regardless of whether a computer's desktop is run from a local PC or a far-flung server in a co-location facility, the power that's required to boot and maintain that operating system has to come from somewhere, as does the energy required to run the access client.
As the number and variety of software- and platform-as-a-service vendors increases, so too does the use and increasing dependency people will have on virtualized machines deployed by those vendors. Companies everywhere are seeking the savings and business agility that come with virtualization, and server density and data center consolidation continue to trend upward.
Except for slight dips after the turn of the century, world spending on servers, power and cooling has continued to rise continuously since 1996, according to IDC. During the period that extends through 2012, the installed base of systems is expected to rise by more than eight times, increasing from around 5 million units to more than 40 million, estimates the researcher.
Click image for full-sized view
For VARs, this trend spells new opportunities to help keep customers focused on power usage efficiency. They will see opportunities as customers focus on increasing power protection to match increases in compute resources, employing services to monitor energy systems to maintain peak efficiency, and integrating power management systems with an organization's existing support and reporting infrastructure.
"One thing that gets forgotten about when deploying virtualization environments is what to do about power," said Jeff Ames, senior business unit manager in the Distributed Power Solutions Division of Eaton. Among the risks of an underpowered data center are tripped breakers, overloaded circuits and a rise in unpredictable system behavior and data loss due to power anomalies.
Seizing The Power
Power management companies are seizing the opportunity. They're combining the abilities of their own software to control hardware power with the features of the hypervisor. Most far along in this type of integration is Eaton, whose plug-in for VMware puts power management features right inside the vCenter.
Click image for full-sized view.
Eaton's Intelligent Power Manager for vCenter plug-in eliminates the need to switch to a separate tool and maintain separate settings. Host servers are discovered and listed in the center pane, and when selected, can be configured by administrators for automatic actions of hosts based on power events such as shutting down or migrating (using vMotion) to a host that's not in distress. IPM events appear on vCenter's logs and in server dashboards. Eaton's Intelligent Power Protector module can initiate orderly shut downs of ESX and ESXi servers and virtual machines as well as Microsoft Hyper-V and certain versions of Linux.
The CRN Test Center has tested Eaton's Intelligent Power Manager for vCenter plug-in, and liked the convenience of having power management features and logging within the vCenter interface. To test its integration with vMotion, we used IPM to configure a pair of hosts, each with separate power sources and set them to migrate one-to-the-other immediately following a power event. Then we pulled the plug. After 10 seconds, the distressed host entered maintenance mode and began to make its move. A minute or two later, the troubled node had successfully shed its workload to a healthy host and shut down. All events were recorded in vCenter's logs. Hosts can also be set to automatically migrate back to their starting points once power is restored.
IPM can be set up to send power-event alerts and notifications through e-mail with an eye toward helping maximize system availability, simplifying asset management, remotely controlling power to connected devices and performing power analysis and trending reports. In other words, IPM can be the an MSP's on-site field technician.
Next: Seek Duplicitous Redundancy
Redundancy is a key part of keeping those computing assets humming. For VARs seeking power-related consulting income, step one might be to perform a power audit with an eye toward increasing the amount of redundancy in the server room.
Servers are provided with redundant power supplies not just in case one fails, but also to allow the provision of power from different circuits, separate power distribution units, discrete UPSes or all three. Eaton's Ames says availability is kept high by eliminating as many single points of failure as possible. "Not having a single point of failure on any critical infrastructure equipment is paramount to the success of those networks and their deployment of data center and distributed networks," he said.
While you're on site performing your data-center power audit, you might also look at the placement of compute resources and the amount of hot- and cold-air separation. Just as you shut the windows when running the car's air conditioner, so too should hot air be kept separate from the cold air coming in from the ventilation system or chiller units. Where possible, suggest the addition of containment curtains or partitions, and be sure all empty rack space is closed off with blanking plates or foam. Even cardboard sheets or boxes placed between ceilings and rack tops can help improve separation.
The next -- and most lucrative - -part of your power audit is to measure the data center's power usage efficiency, or PUE. This number is derived by dividing the total energy used by all aspects of the data center by the power used by computing resources alone. Among the equipment included in the total energy calculation should be that of the physical infrastructure such as cooling units, UPSes, lighting, fire protection, physical security devices and the like, plus of course, the IT load. This includes mainly networking, servers and storage. Do not include energy used by data center staff computing resources, lighting or environment.
To calculate PUE, simply divide the IT load power usage into and total power usage of the physical data center infrastructure and compare it to national averages. These can be found at the Uptime Institute. Lower numbers are better. A PUE of 2.0 or 2.5 is considered good or average for a medium-sized large data center. The most efficient data centers in the world are run by Google, with a PUE around 1.2. You too can learn and employ Google's power savings methodology yourself and for your customers.
The Power Of Power
The virtualization craze has sparked new growth in the data center for servers, storage and the power and cooling systems to operate them. "Customers are facing problems they never had before," said Philip Fischer, data center business development manager for North America, at APC by Schneider Electric. "They're not experts; they're looking for partners."
Data center growth is not limited to large companies. "A common misnomer about data centers is that they're always large," he said.
In reality, about 80 percent of data center growth takes place within small and medium-sized businesses. "This represents a huge opportunity for helping with changes, adding capacity and segmenting the data center," Fischer said.
As data centers continue to consolidate, the need for increased power and cooling will become apparent, either by accident or by design. With just a small amount of specialized knowledge, resellers can benefit either way.