Trend

Cooler’s better

A one-degree drop in temperature can help a data center save 4% on cooling. Akhtar Pasha looks at technologies that can help conserve power

In the context of power and energy consumption, doing more with less is the new diktat for the data center (DC). In the past, more DC performance was the goal. Power and cooling costs were just the price you paid for it. As energy costs skyrocketed, processor and memory technologies have made performance the abundant resource, while power and cooling are at a premium. As server density rises, so do power requirements. As power increases, so does the heat output, and already, the inability to power and cool a DC effectively is preventing many companies from achieving their IT goals.

It takes a comprehensive strategy to scale DC capabilities, while simultaneously containing power and cooling costs. With the increasing deployment of high-density racks of equipment, the greatest challenge for existing DC becomes finding ways to extend the life and capacity of the DC with limited cooling resources.

The deployment of high-density racks of equipment is creating power and cooling challenges for DC worldwide. The server densification trend is intended to create efficiencies in floor space, cabling, and systems management. However, the growth in power density (watts per U) with each new server generation is causing DC to limit rack utilization based on cooling capacity. They are in dire need of new cooling solutions to reap the benefits of dense server configurations.

Density accelerates the issue

Power and cooling are issues regardless of form factor. However, increased server and processor density has accelerated the demands. Today, DC power density is increasing an average of 15-20 percent a year. Customers have started using blade servers to simply some of the first to recognize critical power and cooling issues due to higher physical density.

Paying for power

Over the last decade, server power costs have more than doubled, placing greater stress on cooling, and power infrastructures. According to experts, average server power densities have increased ten times over the course of the past decade. DCs are seeing annual expenses—just for cooling—reaching into the multi-million dollar range. Therefore, every watt conserved provides real cost savings back to the business.
Karthik Ramarao, Director-Technology Systems Practice, Sun Microsystems India said, “Today’s server processors are designed to reduce the power envelope by 10-20%.” He argued that it wasn’t just servers per se but the fact that businesses acquired systems willy-nilly and deployed them whenever there was a new application or project all of which contributed to rising power and cooling costs.

There are no easy solutions. Rajesh Dhar, Country Manager, Industry Standard Servers, Technology Solutions Group, HP India, said, “Although there is no quick fix, multiple technologies are available, which can help control costs. We call these the Chip to Chiller technology.” Dhar said that there are various technologies that can be deployed from chips [processor] to chillers.

Focused cooling to avoid waste

• Blades vs. Racks: Blades can reduce the consumption of power but suffer from one shortcoming which is that they are densely packaged resulting in an increase in the amount of heat generated. So much so that if not cooled adequately this could end up frying your processors or melting other parts. Ramarao said, “Today a rack mounted server can give that performance required for a critical application and the way they are designed can reduce power and cooling costs. These are designed to work on differential temperatures in such a way [air flow inlets and outlets] that it runs at 15 degrees.”
• Virtualization means fewer boxes to cool: Dhar said, “Servers are one of the principal culprits for increasing the power and cooling costs in a DC. So go to the root cause and pick and choose the architecture that can reduce DC operation costs.” There are solutions that can help businesses virtualize their servers [using VMware on x86 platforms, for instance]. You can also virtualize applications using the OS. For example, Sun Solaris, IBM AIX, or HP UXi help partition a servers to run different applications on the same box. For example, a 4P rack server can be artitioned/virtualized to run ERP, database, CRM, HRMS or any other application using these proprietary UNIX server OSs.

Similarly, you can also virtualize storage using networked storage, networks infrastructure and even applications. These virtualization technologies have to be fabricated into the DC design.

Dhar continued, “It is vital to address the power and cooling aspect of the DC and businesses should look at targeting three key areas as to what is going into servers, racks and the DC. Most DC managers do not attack these three areas. It has to be addressed holistically and not in isolation. These small things get trivialized.”

He cited an example of using small factor disk drives that consume 30 watts, cost 10-15% more but over a long period, the cost can be amortized, and it pays itself back. He added, “Pay attention to what you are buying [products]. While negotiating product price if you spend an extra 20 minutes and ask the vendor about the product’s thermal print—how much power it consumes and its efficiency and like, it will help you in long run.”

Liquid cooling

Liquid cooling is not new; it has been used from the days of the mainframe. Today many a large DC is looking at including liquid cooling as an option as they are running out of room to ventilate racks. Improved energy efficiency is just one of the many benefits that this technology brings to the table and it is a lot easier to pump liquid, than air, to where it is needed and it is efficient for denser environments. In this context chilled-water, cooling solutions represent the way forward, and businesses should plan to install them within the next few years. Liquid cooling comes at a lesser cost when compared to air conditioners, as [the latter] requires huge blowers, which add to the cost. This technology is widely endorsed by IBM, HP, Rittal, APC-MGE, and Emerson.

In a chilled water system, an external air conditioner cools water typically between 40°F and 45°F (4.4°C and 7.2°C). This chilled water is pumped throughout the building for use by air-to-liquid heat exchangers. For example, Sun Microsystems in Bangalore is using liquid cooling for its DC in a big way and the company has adopted a solution from APC that is based on chilled water. HP India would be signing a large deal with a large R&D firm soon to implement chilled water-cooling for the DC. If you already have chiller for air-conditioning it becomes easier to use and implement for the entire DC. However, there are three key basic requirements for chilled water-cooling:

• Direct connection to the building’s chilled water system
• A dedicated chilled water system
• A water-to-water heat exchanger unit connected to a chilled water or building water system

Dhar said, “Our new cooling technology such as HP Modular Cooling System (MCS) makes possible the deployment of up to 30 kWh in a single rack, bringing together hardware densities and power consumption levels that have been difficult, if not impossible, to cool previously.” He added that this technology would suite to businesses that are already using chilled water for air conditioning or setting up a new DC. IBM uses Cool Blue technology, which consists of a radiator-like component that attaches to most conventional server racks.

Dynamic Smart Cooling (DSC) from HP, a software-based, air-conditioning-management system that automatically adjusts air distribution based on changing IT equipment loads and heat gain also looks like a practically deployable solution. It uses a network of sensors located throughout the data center facility; the system detects temperature changes resulting from variable workloads and redirects chilled air automatically throughout the raised-floor area. The ‘smart’ piece is the system’s use of computational fluid dynamics (CFD) modeling to understand the flow of air within the data center and route cold air to hot servers.

We recommend that DCs use focused cooling technologies to avoid the wastage of power and cooling. What is required is a holistic approach to solve the issue and there is no quick fix solution ready.

akhtar.pasha@expressindia.com