Industry Outlook is a regular Data Center Journal Q&A series that presents expert views on market trends, technologies and other issues relevant to data centers and IT.
This week, Industry Outlook asks Clemens Pfeiffer about the role of software-defined power in data centers, particularly in light of the trend toward software-defined everything. Clemens is CTO of Power Assure. He is a 25-year veteran of the software industry, where he has held leadership roles in process modeling and automation, software-architecture and database design, and data center management and optimization technologies.
Industry Outlook: Briefly, how would you describe a software-defined data center compared with a traditional data center?
Clemens Pfeiffer: In a software-defined data center (“SDDC”) —a term heavily promoted by VMware—compute, storage and network resources are all virtualized, abstracted away and delivered as a service. SDDC enables dynamic provisioning and ongoing management of all IT resources without taking applications down for upgrades or infrastructure maintenance. This, in turn, not only improves reliability, but it also helps with efficiency and scalability of IT services.
Although application abstraction has eliminated a lot of IT infrastructure-related issues that historically caused application downtime, power on the other hand now becomes the dominant reason for application outages in a data center, as none of the current software-defined solutions addresses or abstracts applications from power dependencies. Independence from power-related issues is still currently handled with multiple locations and multiple-redundant power infrastructure, relying on failover or disaster recovery systems to react to problems. But the picture does not look as good as spending levels would indicate. Despite all the investment, application failover to another site most often requires manual intervention before it successfully completes.
IO: The software-defined data center is capturing a lot of attention, but little is said about power and cooling in this context. What’s the story here?
CP: The reason is actually fairly simple: data centers today are managed by two groups with different responsibilities. Facilities manages the building, security, power, cooling and all the infrastructure needed for IT people to provision their equipment. IT personnel assume that when there is space in a rack and outlets are available, they can put more servers or other IT equipment into this rack. So for these IT people, a data center is mainly servers, networking, storage and applications—they don’t consider power and cooling part of their world. Because the software-defined data center has been defined by IT people (mainly VMware), with others picking it up, the ignorance of power and cooling or even the building is expected. Unfortunately, incident statistics show that this ignorance now poses the biggest threat to application availability, as any abstraction from storage, networking and servers will be unable to keep the application online when power fails.
IO: Data center infrastructure management (DCIM) is an important element of data center efficiency. How does it relate to the software-defined data center?
CP: Although the SDDC takes a purely IT-centric view of the data center, facilities people historically had a building management system and other components that they had to manage; you can think of it as data center infrastructure. If you want an integrated management environment for this infrastructure, you get what is called DCIM—something that facilities people are looking for as a mechanism to help with efficiency improvements and management of the data center. With all the talk about “software defined,” some now even call DCIM software-defined infrastructure. But we still have to consider the operational separation of facility and IT infrastructure, something that poses a potential risk to the application availability. Obviously for someone who looks at IT and facility components as an integrated solution, a “software-defined data center” should also include both, but historically that’s not the case. And worse, DCIM or software-defined infrastructure has no relation to the application today, which is the main reason the data center exists in the first place.
But even in DCIM or software-defined infrastructure, the dynamic management of power and its association with the application falls short or is nonexistent. As a result, it won’t fix the limitations of the software-defined data center and it won’t help to abstract the application from the power infrastructure.
IO: What is “software-defined power” and how does it relate to the software-defined data center and DCIM?
CP: With neither DCIM nor SDDC addressing the abstraction from power, an additional requirement must be met: that is what we call software-defined power, or the abstraction of the application from a specific power source. This task is done by moving the application automatically from one data center to another or one data source to another, effectively freeing it from a specific power source by exploiting the multiple locations, data centers and redundancy hardware that has been provisioned for disaster recovery purposes. But this process makes it a matter of routine, automated, tested and ongoing capability that is easily managed from a single integrated dashboard. When done properly, it helps actively manage the application, eliminating potential outage risks from power problems, by shifting the application from one location to another during maintenance procedures, when power outages may occur because of weather or when utility grid issues arise. But it also creates another opportunity that is currently unused by data centers: participation in the lucrative demand-response and ancillary-services market.
Furthermore, keeping spare or redundant equipment in a defined state (e.g., turned off) until needed will save a lot of power (close to 50% in most cases) and will enable use of such capacity for any application as needed, rather than being solely allocated to a specific application all the time.
Obviously, data centers that are already software defined and have a DCIM solution will have an easier time implementing software-defined power than data centers with applications on single servers or mainframes. The abstraction of the application from the IT hardware and the dynamic nature of infrastructure managed using a DCIM system will make it easier to adjust all components related to an application as it is moved from one location to another, because it is already independent of a specific set of IT hardware. The automation used as part of software-defined power, however, needs to integrate with all the other DCIM and software-defined components to seamlessly make changes as needed and thus keep the applications up and available all the time—something Power Assure has done for years.
IO: How does software-defined power maximize application uptime?
CP: With power being the primary reason for application outages, solutions that address the issues surrounding power will therefore maximize application uptime. Software-defined power is certainly the most promising and cost-efficient one and will beat any hardware redundancy from a cost/benefit perspective. Although hardware-based solutions can provide 2N+x redundancy that looks good on paper, however, it does not address the dependency of the application on a physical location, nor does it shield applications from failures at the grid level, be they weather or operator related.
The only way to actively move applications out of a potential danger area is to have a “software-defined power” solution that does exactly that—automating the transition of an application from one location to another and ultimately turning off and on the associated IT and facilities equipment.
In addition to increasing availability by affording greater immunity from unplanned downtime caused by volatile power sources, shifting application workloads across data centers also makes it easier to schedule the planned downtime needed for routine maintenance and upgrades in each data center. Together, these improvements have the effect of maximizing application uptime with absolutely no adverse effect on service levels, performance or quality of service.
IO: What benefits can holistic allocation of IT resources in and across data centers yield, and how do those benefits relate to software-defined power?
CP: First, allocating IT resources across multiple data centers and allowing the application to move around as a matter of routine will increase the reliability and availability of the application, as you can move it proactively rather than reactively when bad things already happened.
But when you can also turn off and on equipment dynamically as part of the processes, you can shift and control power consumption across different locations, opening up additional contract opportunities for your energy purchases. You can participate in demand-response and other ancillary services. You can do peak shaving, adjust consumption against prepurchased limits and avoid single-spike demand charges. We see up to 50% savings from dynamic capacity management and up to another 25% on energy-market incentives, as well as follow-the-moon pricing arbitrage and other programs. If you also use more-sophisticated forecast models, you can even do real-time and day-ahead energy purchase contracts, avoid peak-pricing windows and, in general, become more flexible in how you procure power.
IO: What steps do you recommend companies take to implement a complete software-defined environment covering all elements of the data center, including servers, network, storage, cooling and power?
CP: Software-defined power requires IT to continuously match resources with application demand and to shift an application to the data center with the most reliable and cost-efficient power source at any given time within the limits of application service-level guarantees.
Software-defined power works in enterprise data centers by following the five implementation stages below:
Get a baseline of each application, its utilization and the associated power consumption and backup/disaster recovery capacity. Gain real-time insight into power consumption and IT utilization.
Analyze data center efficiency using the baseline information and real-time measurements to identify inefficient or underutilized IT equipment.
Automate dynamic application-workload movements across multiple data centers and adjust associated facility infrastructure dynamically using existing standard operating procedures.
Make dynamic workload management and server capacity adjustments a matter of routine to not only react to catastrophic events but also act before they happen.
Integrate with energy-market intelligence to identify cost, demand-response and ancillary service opportunities to monetize on the dynamic workload management across data centers, and capitalize on energy-market participation to fund the implementation of software-defined power.
The post Industry Outlook: Software-Defined Power appeared first on The Data Center Journal.