Bed sheets cover up widening data centre crisis

A Sydney data centre has resorted to hanging cotton sheets from the roof to alter in-room airflows, highlighting the cooling crisis faced by many Australian operators.

It is understood that other centres are placing chimneys over hot equipment to exhaust the hot air straight through the roof, or installing false roofs between racks to negate inadequate floor pressure.

“We see a lot of inefficient use of air-conditioning,” said Brad Ferguson, managing director of The Frame Group, a Sydney-based technology services company.

“There is no regard to the aerodynamics or airflow within the facility or even how equipment is placed into racks. We’re currently working on one [centre] in Victoria like this.”

The ‘quick fixes’ underlie what appears to be a growing trend in the data centre space – that many businesses have outgrown their buildings and aren’t sure where to turn next.

“A lot of customers are taking out short-term leases because they don’t know if the facility can handle their requirements long-term,” said Bruce McEwen, director of Galileo Connect.

“The only reason to go short-term is if you want to move, but the cost of moving a data centre is horrendous and failure isn’t an option.

“We’re seeing customers staying in older, unsuitable facilities due to the cost and dangers of moving out,” McEwen told iTnews.

Those who stay are implementing strategies such as those described above to extend the life of their buildings.

Others are playing what McEwen calls ‘hot/cold games’, referring to use of the hot/cold aisle concept to prolong the capacity of an existing centre.

This involves dividing the centre into a series of ‘hot’ aisles where racks face each other, and ‘cold’ aisles which run behind each row of racks.

Hypothetically, cold air pumped into the cold aisle is drawn through the fans in the mounted equipment on both sides of the aisle. The resulting hot air is exhausted into the adjacent aisle.

“Hot/cold design causes unnatural placement of equipment,” said McEwen.

“A very wide data hall is just an absolute basket case to manage from an airflow point of view. If you don’t have to play hot/cold games then that’s a godsend for design.”

McEwen believes the mid- to long-term resolution is that new energy efficient, flexible and upgradeable data centres need to be built and customers moved into them ‘in a planned and scheduled way’.

This modular approach to data centre engineering and design is critical because the next generation of centres need to provide a safe infrastructure environment for the next 20 to 25 years.

Read on to page 2 for an alternative strategy to cotton sheets.

Global broadband revenues to decline, analysts predict

Despite a growing demand for enterprise broadband access, falling prices could mean a decline in global broadband revenues during the next five years, analysts say.

According to analyst and consulting firm, Ovum, enterprise uptake of Voice over IP (VoIP), network-based applications and next-generation technologies are expected to drive demand for broadband access.

Ovum expects broadband connections to grow from 59.8 billion connections globally in 2007 to 87.6 billion in 2012, representing a compound annual growth rate (CAGR) of 8 percent.

However, subscriber growth will be insufficient to support revenues through an overall decline in broadband access pricing during the next five years.

Enterprise demand is expected to grow revenues from $59.4 billion in 2007 to $62.4 billion in 2009. However, revenues are expected to decline to $55.6 billion in 2012.

Analysts expect the decline in revenue to be driven by high competition and falling broadband prices in developing markets.

According to Melbourne-based Ovum analyst Claudio Castelli, the developing markets of China and India will overtake Western Europe to become the region with the largest rate of adoption of DSL.

“We see high competition for services and continued pressure on DSL pricing; however, demand is growing and counteracts this revenue loss,” he said.

“We forecast revenue increases in the China/India region with revenues rising from $18.1 billion to $23.1 billion in 2012, which will be more than 40% of the global market,” he said.

IPv6 doomed, requires act of God

ISPs and carriers have spectacularly failed with IPv6, to the point where they need to resurrect network address translation (NATs) to persist with IPv4, a leading expert has warned.

In what at times resembled a doomsday prophecy, Asia-Pacific Network Information Centre (APNIC) chief scientist, Geoff Huston, told a packed AusNOG conference that ‘it’s way too late’ for IPv6 and that the industry will need ‘religion’ to get it working before the impending exhaustion of IPv4 addresses.

It is anticipated IPv4 address allocation could now run out as early as January 2009 – particularly as sectors of the industry resort to ‘panic buying’ of addresses as the shortage looms, according to Huston.

“When the IPv4 supply runs out, you’re going to have to pay [to get addresses], and after another ten years the price is going be amazing,” said Huston.

Huston was repeatedly critical of industry inaction around IPv6 deployment.

“You’ve implicitly selected the failure option [for IPv6],” said Huston.

“You haven’t done anything with IPv6, and now it’s way too late.”

When IPv6 was conceived, it was anticipated that networks would be ‘dual-stacked’ to run both IPv4 and IPv6 for a transition period.

Once IPv6 was ubiquitous, IPv4-based functionality would be switched off, heralding the next generation, according to Huston.

“It was all meant to be finished now, but we haven’t started,” said Huston.

“There’s no IPv6 out there in production land and no IPv4 addresses left. If you want IPv6 to happen now, some external factor is going to have to kick you in the pants.”

Failure, Huston said, is an option with IPv6. Rather than continue to look at IPv6, he encouraged the industry to persist with IPv4 with ‘intensive’ use of carrier-grade NATs.

“When addressing gets really expensive, it’s not obvious that you’ll jump to IPv6 – it’s more likely that you’ll NAT very intensively,” explained Huston.

This is because any IPv6 rollout will be funded out of ISPs’ own ‘marginal revenues’ rather than by customers who, by and large, can’t see the benefit of paying more to migrate from IPv4 to IPv6-based networks, Huston said.

“Customers will not pay,” explained Huston, “because while IPv6 might be technically whiz bang, it represents the benefit of a minor technology upgrade at the cost of a major forklift upgrade.”

Pushing a NAT IP address hard could result in a capacity of 200 hosts per address – and a potential billion customers per ‘slash eight’.

Slash eights are class A networks that have an 8-bit network ID.

However, the suggestion was questioned by attending members of the IPv6 Forum, who claimed the money spent on NATs would be better invested in furthering IPv6 deployments.

But according to Huston, the argument is no longer over to NAT or not to NAT.

“It’s more about questions like how many NATs can I deploy,” said Huston.

The AusNOG 02 conference concludes in Sydney today.