Arthur Cole spoke with Clemens Pfeiffer, CTO of Power Assure.
Cole: The failure of Cassatt Corp. was a bit of a surprise considering the company had an innovative, if somewhat complex, solution for a serious IT concern -- power management. What has Power Assure done to keep complexity to a minimum and to foster integration into IT environments without a lot of changes to existing infrastructure?
Pfeiffer: Cassatt developed a very innovative software solution for provisioning virtualized servers, but when they were not generating sufficient sales, they changed their strategy and moved into the Power Management market. However, their product was not designed for power management and had some shortcomings. For example, selling an enterprise software license to build a policy-based custom solution requires a lot of upfront investment in license fees and development time. It also requires long-term maintenance and is limited to the policies and rules someone can implement.
"To have a viable power management system, you have to take a holistic view of the data center rather than focus on one specific part of the system. Otherwise, you will make the wrong decisions."
With unpredictable events happening all the time, along with ongoing changes of underlying applications, the result is a never-ending development cycle without achievement of significant savings. Limitations in scope and target environments and lack of integration with facilities systems made Cassatt's job even more difficult.
Power Assure takes a fundamentally different approach. We designed our system from the ground up to specifically address the need for automating power management in data centers. First of all, our software is delivered as a service. Customers don't have to do any development. Power Assure does the set up, installation and initial configuration and then charges as customers actually use the product. Second, our system is not rules- or policy-based. Our software makes decisions dynamically, in real time, according to minute-by-minute changes in customer demand. Finally, Power Assure uses a business process automation platform to implement operating procedures across existing applications - the way operators or system managers would perform the tasks - to adjust capacity to actual application demand. This includes servers, network elements, power and cooling components.
As a result, Power Assure supports not only virtualized environments but also load-balanced Web server farms, cluster, cloud, Citrix and other large server farms. By integrating with the specific application management environment, Power Assure complements existing investments and leverages maintenance procedures of these environments to minimize its own maintenance. We don't require customers to replace their management systems: no forklift upgrades, no client software on their servers. This "minimal intrusion" approach is very appealing to customers.
Cole: When you think of power management, you naturally think of shutting down idle servers. What are some of the other components in a power management system?
Pfeiffer: While most of the savings come from shutting down idle servers or putting them to sleep, there are other components that need to be factored in. First, servers need to be taken out of allocation for each individual application. Load balancing may need to be adjusted; routing may need to be changed; site-to-site load shifting might be required; and cooling might need to be adjusted as servers are turned on or off. Therefore, Power Assure takes a holistic approach and integrates with the customer's existing systems - cooling, network, power and server equipment - to allow for an unprecedented level of flexibility when configuring the operating processes. This includes monitoring and management of power from the utility feed to the individual outlet, temperature, humidity, server and network utilization as well as cooling capacity. To have a viable power management system, you have to take a holistic view of the data center rather than focus on one specific part of the system. Otherwise, you will make the wrong decisions.
Cole: Networking devices also draw a lot of power, but they tend to fall off the radar at firms that are concentrating on their server, storage and PC infrastructures. Any chance that we could see things like switches, routers, HBAs and the like brought into the power management fold?
Pfeiffer: Absolutely. We are already doing some of that today. We incorporate servers, network elements and cooling into our environment. The areas we do not address at this point are storage and desktop environments. There are also some inherent limitations when it comes to turning off networking gear. While network adjustments can be made, most data centers do not support on/off capabilities on the network equipment as that would limit access to the outlets/meters within the proximity of such network equipment.
Nevertheless, we have customers who, in their test labs, would like to turn on/off network equipment, or dynamically reconfigure such equipment. We are currently developing support for this capability into our platform. Long-term protocol extensions like EnergyWise can enable more granularity which, along with IPMI-based turning on/off server components, is part of our near-term roadmap. The same is true for storage. While at this point the dynamic management capabilities are limited, some new virtualized storage solutions are in the works that will allow us to integrate them as well.