Advanced Computing in the Age of AI | Thursday, March 28, 2024

Hands-Off IT Power Management 

IT power management is the low-hanging fruit of green computing. DhaaniStar is a network-based, adaptive management solution that strives to maximize both power savings and user satisfaction with a policy-free approach. 

PC power management is the low-hanging fruit of green computing. A 2009 report found that US businesses waste about $2.8 billion annually to power unused desktop computers, creating 20 million tons of carbon dioxide emissions. By powering off unused devices, or switching machines to a low-power state (like sleep or hibernate), businesses can significantly reduce their power bill. A centrally-managed power management strategy is the surest way to stopper this energy drain.

Corporate PC Power management solutions are gaining in popularity. Major brands like Walmart, Boeing, and AT&T are converts. But it's surprising how many businesses haven't yet jumped on this green bandwagon. Many organizations think they can go it alone by setting their own power profiles. Others are concerned about the security or manageability of their IT infrastructure.

A new solution from Dhaani Systems aims to counter these worries by offering the industry's first completely hands-off power management system, DhaaniStar. The network-based, adaptive management appliance was designed to maximize both power savings and user satisfaction. The company cites an energy cost savings of 50-90 percent at various customers and a customer satisfaction rating of 100 percent.

According to CEO Shankar Mukherjee, the company launched 18 months ago with a first generation product that supported management of desktop PCs, laptop PCs, thin clients, and non-virtualized servers. Last quarter they added support for cloud-based, virtualized servers running on VMware. They are currently working on building up sales and support infrastructure all over the world.

"Our mission is to make existing IT equipment more energy efficient without disrupting user productivity or business productivity," says Mukherjee. "There are many solutions available in this market, but most have issues dealing with the 21st century work environment. Hardly anyone has a set schedule anymore. Monday is different than Tuesday and summer is different from winter. In most environments, we find that every user has more than one device and every device has more than one user. Even when there is one main user on a machine, IT is the secondary user."

Mukherjee explains that most maintenance work is done off-hours especially for the fixed assets, the desktops and servers, and it happens remotely. All the fixed endpoints, desktops, PCs and servers are almost always on. At idle, a PC typically consumes about 70 percent of its maximum power, while a server consumes roughly 80 percent of its maximum power. A desktop system with the monitor switched off still draws 70 percent of maximum power.

"We built the industry's first network-based appliance to handle these uneven usage scenarios," says Mukherjee. "All these devices are essentially communicating with each other using the entire network. To manage these devices' power usage without disrupting employee or business productivity, we have come up with an appliance that models the aggregate usage of every device over a period of time, from 48 hours to 7 days, and it comes up with a very accurate short term prediction that essentially says if you find a device idle, determine how long is it likely to remain idle. Has the employee stepped out for a quick coffee break or for a longer lunch meeting or for the rest of the day? Using this approach, not only does the software put the machine in a lower-power mode, it also wakes up the machine in time for its next anticipated use."

From the end-user perspective it is completely transparent. The solution is a completely client-less solution; there are no policies to be set. Dhaani Systems just received a patent on this real-time predictive analytics-based technology.

Of course, no prediction algorithm is 100 percent accurate and certain scenarios are difficult to predict. Mukherjee relates this example: if a user has gone on vacation for a week or two and the software is saving money this whole time, and the user comes back and has to wait for his or her system to boot back up, Dhaani considers this a customer dissatifaction event because the appliance's mission is to turn the computer on before the next anticipated use.

What the software does in this instance is learn from that event and recalculate this machine's usage profile. Once the prediction accuracy reaches a certain minimum threshold, power management on that device is reinstated. Because the software is continually learning and adapting to changing work environment, IT never needs to get involved in managing the solutions.

The appliance sits in the same network as the PCs and servers that are being managed. As long as the devices are behind the same firewall it can be managed by the product. The appliance can be globally deployed for a very large organization but can be and managed centrally from their headquarters. Customers must have at least 50 or more PCs and servers, but there are no upper limits to scalability.

The software can be installed in about 30 minutes and supports any type of hardware so long as it is running a version of Windows, Linux or Mac OSx. There is a secure high-availability option for customers who are seeking those qualities.

Mukherjee says there are about 14-18 other PC power management solutions on the market and they are all client-server solutions. A time-consuming installation process is one potential headache, but the execs that Mukherjee spoke with were more concerned about the configuration process and the possibility of end user complaints. Some products require significant employee involvement and this can interrupt production and result in calls to the IT department.

In case studies, the savings were significant. A computer lab at UC Berkeley saw their average energy cost per year per PC go from $68.20 to $11.30. That's a cost savings of $56.9 and an overall savings of 83.4 percent. A second case study involved a medium-sized semiconductor company in Silicon Valley. With DhaaniStar, the average energy cost per PC went from $150 to $39, a cost savings of $111 per year (a 74 percent reduction). The larger the IT infrastructure, the greater the potential savings. The average time-frame to see a return on investment is a year or less.

The demand has been there for a long-time, but the install base is still pretty small, small enough that Mukherjee considers this a greenfield opportunity: saving energy without disrupting people's work.

Pricing is based on per case basis, with the ability to purchase the product outright or utilize a subscription model which is about 50 percent of the savings on a yearly basis. Most potential customers initially are attracted to the subscription model, but after reviewing the particulars with their CFO wants to purchase the product and not share their savings.

EnterpriseAI