Green Notes

IBM, Syracuse University (SU), and New York State have built a computer data center on the university’s campus that  incorporates advanced infrastructure and smarter computing technologies to make it one of the most energy-efficient data centers in the world. The data center is expected to use 50% less energy than a typical data center today making it one of the greenest computer centers in operation.

A key element of the design is an on-site electrical co-generation system that uses natural gas-fueled microturbine engines to generate all the electricity for the center and provide cooling for the computer servers. SU will manage and analyze the performance of the center, as well as research and develop new data center energy efficiency analysis and modeling tools. The data center operates completely off-grid.

In addition, a liquid cooling system has been created that will use double-effect absorption chillers to convert the exhaust heat from the microturbines into chilled water to cool the data center’s servers, with sufficient excess cooling to handle the needs of an adjacent building. Server racks  incorporate IBM’s Rear Door Heat eXchanger “cooling doors” that use chilled water to remove heat from each rack far more efficiently than conventional room-chilling methods. Sensors  monitor server temperatures and usage to tailor the amount of cooling delivered by each Rear Door Heat eXchanger–further improving efficiency.

The project includes the creation of a direct current (DC) power distribution system. In a typical data center, alternating current (AC) electricity is delivered by a central power plant through the local utility’s electric grid and then converted to DC to power the servers. This conversion process results in power loss. By directly generating DC power on-site, transmission and conversion losses are eliminated.

What makes the IBM-Syracuse University project different from many other similar ventures is the fact that special attention has been paid to greening the actual infrastructure of the data center itself, not just the computer hardware and software. Other design solutions that we recommend at Green Rack Solutions to update your data center’s basic design are:

• Managed airflow system to avoid hot spots, to assess cooling capacity required for IT equipment and to separate cool intake air from warm server exhaust air.
• Using energy saving cooling methods such as airside economizers and direct evaporative cooling.
• Larger but fewer variable speed motors to power air handlers for energy saving.
• Data center designs that eliminate need for raised floor which creates additional losses and leakages in the air handling system, thereby reducing the efficiency of the overall cooling systems.
• Designing by density zones. High-density applications represent 10 percent to 15 percent of a total data center usage. Medium-density apps account for another 20 percent. The rest is low-density. If you mix and match densities you save money on the build-out.

In datacenters, there is a tendency to overexert and do too much if that makes sense. After all we want to make sure that everything is working well and beyond to avoid problems during uptime. Over provisioning and over allocation often lead to overspending in terms of indulgence in excess power and hardware. Apart from virtualization, increasing storage efficiency and equipment efficiency, Power capping is a new technology that datacenters are looking into right now to address this issue. The concept is logical enough as the name suggests. Power capping sets a cap on the power a server can consume at a given time without always leaving allowances for “in case of maximum load or maximum utilization” and is uniformly applicable for all servers.

Servers come in all shapes, sizes and uses. There are many scenarios that determine their usage like customer requirements and nature of work. So they do not all run to full capacity at all times and some maybe even never as many datacenters choose to not deploy their entire arsenal at the same time. So why do we allocate all this power to them without discerning? Why do we set a standard uniform allocation irrespective of functionality? Also more times than not, the suggested power allocation referred to on hardware pieces are much more than is actually needed. Again everyone is trying to be on the safe side and being careful about not interfering with the performance level, but unwittingly wasting more resources in the bargain. That is why power capping makes a whole lot of sense. Without overstepping the power limit, it allows us to save on space and increase your server density by ascertaining that your servers do not use more than their allocated power at once.

Even though many IT managers are hesitant to experiment lest important applications get interrupted, power has proven to be such an expensive and environmentally unfriendly commodity that many datacenters have decided to take their chances on it .Companies like Intel, IBM, Dell and AMD have released power capping management software to facilitate the process. For example Intel’s Dynamic Power Node Manager Technology along with ancillary management software helps servers running on Xeon 5500 chips to vary platform power on a case by case basis without reducing the performance level through monitoring and assessing real-time data of a server platform. The ceiling on the platform power is thus adjusted to your optimal individual power budget . In case it figures that the budget cannot be met with without interfering with the performance, it will alert you on your choices. Intel also has developed a software addition to the Node Manager called Intel Datacenter Manager, designed to monitor and control power for a group of servers.

Baidu, China’s largest search company has reported success recently using Intel’s power-capping technology. In fact the companies released a study that stated that a datacenter using this technology could save up to 40 watts per system without performance impact translating into as much as 20 percent additional datacenter capacity within the same rack-level power envelope, and a potential rack-density improvement of 20 to 40 percent. Prior to investing in power capping technology, Baidu was leasing racks at a datacenter where the company sought to save money by cramming as many machines as possible into the fewest number of racks. Post Intel’s power capping tools, they were able to establish an optimal workload corresponding to maximum power reduction with least performance loss. Then they determined the approximate amount of power the servers needed to just idle. After narrowing down on the precise number of servers per rack , they successfully established a cap at the rack level bringing down the power consumption from 900W to 750W.

It is always better to understand the full implications of power capping before deciding if power capping technologies in full blast would be a viable option for your data center. A trial run before full scale operation is recommended to protect important applications. Please contact us at Green Rack Systems for further consultation.