Green data centers face software blues
Many enterprise software products have not been architected to be inherently green, thus putting a strain on power consumption at data centers with energy-efficient hardware. However, for green data centers to live up to their name, companies will also have to show more interest and invest in green software, say industry watchers.
Gartner Research Vice President John R. Phelps, for one, thinks the challenge of setting up a green data center needs to be seen holistically. It is too simplistic to say that just solving one aspect will resolve the power usage problem in data center environments, he said.
He went on to acknowledge in his e-mail to ZDNet Asia that "it is correct to say a large amount of software is not written with efficiency of the central processing unit (CPU) in mind". Developers, though, are slowing coming around to understand that other programs will be running on one server and that it can make a difference for power usage to make programs "more efficient and have less looping", said Phelps.
Furthermore, with data center virtualization quickly gaining traction, application mobility features are starting to be integrated into software offerings, noted the Gartner analyst.
"We see the move to enable the shifting of application workload on the fly so that underutilized servers may be cleared out during slack time, [which will] enable [the] power-saving mode [to kick in] or allow for the powering down of servers," he said. For this to work, vendors need to look at "providing fast shutdown and start-up of systems", added the analyst.
Phil Hassey, Springboard Research's vice president of services, agrees. He told ZDNet Asia in an e-mail interview that not-so-green software products are "definitely part of the overall issue with data center power management issues". However, he added that the industry tends to oversimplify itself with its push for green data centers and "forget that the core technology and enterprise culture have to change as well".
"Like all fundamental transformations, [green data center transformation] is more complex in the implementation than first thought," said Hassey.
Both analysts were commenting on an earlier New York Times article that discussed the difficulties in creating green data centers and how "energy-indifferent application programming" is one of the key problems today.
The report cited Microsoft's chief environmental strategist, Rob Bernard, as saying that developers are still architecting programs in "the old paradigm", in which such software are allocated too much memory and hold on to the processor for too long.
"A single program that isn't written to go into sleep mode when not in use will drive up power consumption for the entire server. [If] the application isn't energy-aware, it doesn't matter that every other application on the client is," said Bernard in the report.
Springboard's Hassey did point out that the technology is already available for developers to come up with dynamic, self-optimizing software. The problem lies in the desire of organizations and the "ability to implement the requisite change management and cultural redefinitions", but this change will need to be driven by the enterprise, not developers, he added.
This was reiterated by Alex Tay, a service product line executive for IBM Asia-Pacific's site and facilities services group, who attributed the resistance to change to "fear of business disruptions".
"In many cases, companies adopt an 'If it ain't broke, don't fix it' attitude, and this needs to be re-examined," he said. He also went on to add that because of the recent economic recession, companies looking to meet their ROI (return on investment) expectations have held back on a large-scale adoption of such complex software.
As for developers, Tay thinks the decision to create intelligent software that can be configured, optimized and provisioned through automation boils down to "commercial viability". In order to create such software, developers will need to "consider interoperability, standards compliance, security, ease of use and migration", which mean a large investment outlay just on development and testing alone, he noted.
Ultimately, Tay thinks the proliferation of such intelligent software becomes a question of "demand and supply", as companies balance the need to address the many data center issues over time.