Datacentre power technologies rise to the challenge of the blade servers

Uninterruptible power supplies are crucial for the blade age

Rising energy prices and the legislative focus on environmental issues are forcing IT departments to look again at power consumption and efficiency, putting the datacentre firmly in the spotlight.

Research from IDC predicts that by 2008, datacentre users will be spending more on power and cooling systems than on new server technology.

It is important to realise that it is not only servers that consume electricity. Between them, the ancillary technologies used in the datacentre - air conditioning, storage devices, networking controllers and, not least, uninterruptible power supplies (UPS) - can consume just as much power as the actual servers themselves.

UPS are arguably the mission-critical ancillary technology at the heart any business continuity and disaster recovery . They provide emergency power in the event primary power outages and protect the core an organisation's IT systems against power surges.

David Bond, managing director UK specialist Uninterruptible Power Supplies Ltd, says, "At the end the day, UPS systems just supply electrical load. But by far the nastiest loads UPS have historically had to deal with are blades.

The increased power density blade servers has exposed UPS systems that cannot cope. Now there is not an IT professional in this market who would think of running IT systems without an UPS."

Bond says users became aware UPS problems three years ago with the growth ultra-high-density racks blade servers, which deliver more processing power in less rack space, simplifying cabling and reducing power consumption.

But as more equipment greater density is packed into a small space, temperature rises up to 10 degrees are not uncommon, and heat is the biggest threat to UPS battery life. As a result, heat density is becoming a big issue in datacentres.

The largest and most mission-critical datacentres specify power allocation and corresponding UPS safeguards to meet stringent redundancy requirements that add to power consumption and its regulation.

Redundancy is typically achieved though either an N+1 or a 1+1 design. In an N+1 design, also known as a parallel-redundant system, multiple UPS modules are sized so that there are enough to power the connected equipment (N), plus one additional module for redundancy (+1).

During normal operation, the load is shared equally across all modules. If a single module fails or is taken offline for service, the UPS can continue to power connected systems.

In a 1+1 design, two UPS modules are sized so that either is capable of carrying the entire load, which delivers improved availability over N+1 systems. Some global enterprises even insist on a 2N+1 datacentre architecture for complete peace of mind.

Rakesh Kumar, research vice-president at analyst firm Gartner, says the power needed to drive a rack of high-density server blades can be 10 to 15 times higher than in a traditional server environment. He suggests that organisations carefully evaluate the power protection requirements if they need to use the latest servers.

"The best answer to the power and cooling issues is greater design focus at the component level, coupled with the ability to manage power at a high level," Kumar says.

"It will eventually be possible to implement systems that can operate within targeted power and thermal limits, although many of the components that will enable this kind of power management will not arrive in the next five years."

UPS and back-up generators are increasingly designed to manage the electrical noise degradation caused by the ever growing number of smaller and smaller microprocessors and semiconductors contained within servers.

Electrical noise can be caused by utility grid switching, as well as common office equipment powered by mains electricity such as lifts and photocopiers, and is the source of constant power line disturbances.

If the automatic switching function between the UPS system, back-up generators and utility grid source is not properly configured, any lag can disrupt UPS to computer systems and cause data loss, system lock-up and component failure.

UPS unit capacities have steadily increased. Manufacturer New Wave launched a 40kW unit last year. By comparison, APC offers 66.7kVA (kilovolt amps - another way of rating maximum UPS load) power modules, scalable to 1.6MVA.

New Wave claims its "transformerless" model is superior to more traditional UPS designs. Surge suppression and filtering features rely on in-built transformers, which themselves need power to convert mains AC electricity to DC and back to AC again.

For example, a line-interactive UPS replaces the separate battery charger, inverter and source selection switch of basic units with a combined inverter/converter, which both charges the battery and converts its energy to AC for the output as required.

AC utility electricity is still the primary power source, and the battery is secondary. But when utility power is operating, the inverter/converter charges the battery, operating in reverse if the power fails.

Supplier Liebert recommends users steer clear of line-interactive UPS systems for mission-critical applications. Liebert claims that line-interactive UPS units are generally only 85% effective against power anomalies, and have less protection against surges, harmonics and frequency variations.

In high-speed networks, some Liebert customers have found that in going to battery, line-interactive UPS systems drop loads, resulting in data loss. The use of double-conversion online UPS systems ensures continual power conditioning and maximum protection against even daily power anomalies.

Villazim Xhiha, chief executive of New Wave, says online, transformerless UPS technologies give a lot of advantages in terms of power efficiency, management, size and compactness.

"They generate less harmonic noise disturbance that can affect other power consumption in the datacentre. And our remote monitoring facility which goes live early this year will supply added peace of mind," Xhiha says.

New Wave has begun fitting some of its UPS units with in-built, networked power monitoring tools. When connected to a customer's own management systems or those hosted by the supplier, these tools can provide proactive maintenance of power fluctuations and switching capacity.

APC is a major supplier of UPS equipment in the UK (along with Uninterruptible Power Supplies Ltd, Cholride, Gutor and Liebert), and argues that the technology already exists to provide near-100% availability of computing power.

The company claims that the key to fine-tuning UPS power requirements is maintaining clear datacentre standards to achieve predictable behaviour with high-density loads. APC calls for clearer definitions of datacentre power requirements to help with the design and installation of power and cooling equipment to ensure UPS capacity is right-sized and maximises electrical efficiency.

Paul Tyrer, APC managing director for UK and Ireland, says traditional datacentres have been designed with only 50% extra capacity built in to provide room for 10 years' worth of added demand.

"IT and facilities must be able to factor into their delivery requirements about a 14% increase in datacentre power each year," Tyrer says. "But that adds up to a compound growth rate well above the maximum capacities built into many traditional datacentres.

"We are seeing customers come to us to talk about UPS much earlier in the datacentre specification process nowadays as a result. We tell our customers not to design for year 10 any more, but to build capacity in for the here and now because the modular power and cooling products available in the market today are easily scalable and flex to growing demand, allowing organisations to deploy technology as they grow and not 'just in case'."

The modular evolution of UPS units follows the trend towards scalable blades that can be quickly hot-swapped or added as more computing capacity is needed.

Increasingly, just blades are held in purpose-built housing or racks that contain all the connectivity and networking components required, so the trend within power and cooling has been to build corresponding frames around these blade racks to control and monitor heat with inbuilt air conditioning, as well as to house the UPS units and batteries for greater redundancy.

Tyrer says APC has been physically housing modular UPS hardware in the same rack space as the servers they protect for eight years - since the its Symmetra unit.

Nowadays, the supplier is putting its weight behind its Infrastruxure range, which organises access, cables, power and cooling all in one unit around the server farm.

"While the industry tends to look at server statistics in the datacentre, the attach rate such modular power protection products is high, as disaster recovery and business continuity plans come under consumption and efficiency focus," Tyrer says.

There are other simple housekeeping tasks a datacentre manager can usefully perform, such as making sure UPS batteries are properly charged and alternative power sources are on assured standby.

But ultimately, says Chris Dean, managing director datacentre consultancy DMW, businesses need to test the UPS component their computing infrastructure for the future.

"They need to 'pull the plug' to make sure their business continuity works. It has to be done very carefully and takes a lot planning but it is the only way to ensure you are protected. At the same time, businesses need to look at how they can conserve energy used by IT," says Dean.

IT managers can do this from the ground up, by looking at environmentally friendly ways to build the datacentre.

Judging by how key UPS requirements are in today's IT environments, a business may pay a high price for overlooking it.


More information:

Datacentres losing their cool

How to avoid datacentre downtime


Comment on this article: [email protected]

Read more on Server hardware