Some say there are concerns about the energy costs of running these increasingly dense and power-hungry data centers. On the other hand there are a lot of interesting ideas about getting IT green by implementing technical measures. But if it is so easy, what is the reason IT is not green at this moment? The reason IT is not green at this moment is at least as much due to economic reasons.
Who is responsible, who feels the pain and take action? What is your opinion about who cares most about enterprise IT energy efficiency? The CFO, the CIO, IT Operations Management, or others?
This question got quite some reactions on a discussion group on LinkedIn, some impressions of this discussion …
Part of the problem is that many people take the electrical power supply for information processing for granted. But information processing (thus data centers) are part of a complex electrical power supply chain. People are mostly not aware of the tremendous energy losses in this supply chain. Saving a kiloWatt at the end of this power supply chain saves you and thus also society a lot in the upstream of the power value chain. Data centers are significant consumers of energy and there is a large focus on reducing the PUE. It is also increasingly clear that there is much room for improved energy efficiency, with most servers operating under the ten per cent average utilization. The cost of power is not irrelevant, it seems these costs are hidden for many people. Given the reactions it looks like there is still a need to create more awareness about energy efficiency.
- “Without a doubt the single largest component of the power entering a data center should be the IT equipment. Sadly this equipment is not well managed. In fact if we ran our households like we run our data centers we would all be going BROKE!”
- “For starters on the IT side some are turning to ENERGY STAR or other rating systems to find the devices that have the highest performance per watt AND lowest Idle power characteristics knowing full well that the vast majority of the time the devices are running at or very near idle.”
- “What is really shocking is that people still need to be convinced that server OPEX exceeds CAPEX after having been validated over and over for many years now. What’s also really shocking is blade server purchase for technology’s sake rather than for meeting known requirements.”
According to other people the impact of the amount and the growth rate of data storage on energy usage is underestimated, ” … but let’s take a large Data Centre with 100,000 servers. On top of the servers and let’s say those require 15Mw per year, there is an additional 50-70% for cooling and then there is the storage itself. Let’s assume as an average, the data centre has an average of 10 drives per server, 1 million Disk Drives, with an average hourly consumption of 6.5w per drive at $0.15 per Kw/Hr. It should not be hard for anyone to work out how much the data centre will cost to run.”
Questions are also raised about the awareness of component manufacturers, “They (component manufacturers) are focused only on increasing ariel densities because users seem to think that “the cost of power is irrelevant”. (a quote from a component company CEO I spoke to recently).”
And last but not least there is the observation that power transmission networks can be a bottle neck. The power grid can be thought of as a highway system: in most places and at most times, traffic flows smoothly, but at certain critical locations, also known as Critical Areas for Transmission Congestion, there is insufficient capacity to meet the demand now or in nearby future. “For data centre in some European centres, power supply is going to become an issue, we already see power capacity sold up to a decade ahead of initial supply in London, Canary Wharf gets it’s next 4.5Mva feed in 2018, it’s already all sold ahead of supply, mostly to clients with data centres. Halving the amount of power to drives will not solve the problems of supply but it will reduce bills and extend the window at which many very large data centres may need to consider building new capacity with a power station of some sort to provide continuity of supply and the ability to off-set costs by supplying excess energy back to the Grid.”
One recurrent theme was that decisions are made in ‘splendid isolation’. Responsibility of using a data center effectively and efficiently falls across projects, IT departments, business departments, corporate real-estate and last but not least (depending on the sourcing model being used) other IT organizations. “What is really funny is that this is an industry where Green technology will actually save lots of money and so it is really crazy that we continue this way.” With isolated decision making, the measurement and accountability issues and the absence of true energy usage and cost analysis, energy inefficiency becomes the rule.
Repeatedly the observation was made of the gap between the IT staff and Facility staff, those who work on the site infrastructure. “One of the key issues in delaying the migration to more energy efficient data centers (both from an asset point as well as use policies) – that of who owns the power bill. Until the historical (hysterical?) disconnect between facilities/corporate real estate and the IT organization is at a minimum bridged via cost charge-backs, there is little incentive for the IT organization to significantly change their current practice.” Or put it another way, “It is even more of an issue when you consider the real rift that exists between most IT and Facilities departments today. Some of them really don’t get on at all.”
There is also a gap between the supply and the demand site, “I am surprised that more has not been made out of customer demands”.
This current way of working leads to a suboptimal mix It looks like a proper integrated architecture and design is needed to correct unnecessary redundancy and inappropriate use of infrastructure.
Accountability or Who pays the bill
The opinion on accountability is divers:
- Business operation management: “Power consumption is indeed a cost factor. However power consumption itself means little. It should be seen as part of the whole IT infrastructure, and thus, from a cost point of view, be part of the IT side of business.”
- Facility: “In a large number of Companies it would be the facilities manager. Unfortunately the electrical costs tend to be rolled into the building electrical budgets and can get lost if not broken out. This of course is a well documented issue as it can hide the cost from the CFO and CIO. The problem is simple, the person that runs the IT or Facilities doesn’t own the power bill and hasn’t been given genuine responsibility to reduce the power bill in most organizations. In fact, the staff responsible for Capex and Opex have never met in most cases.”
- CIO: “Unless you’re in the automation business 🙂 IT by itself is not a goal, but simply a cost point. It enables you do to business quicker, more cost efficient, etc. To do so, a company invests in and suffers the cost of operating those IT systems. If the results are not worth it, that company would not spend the money on it. So hosting that IT, and the power it consumes, is simply part of the operational costs for owning and running that IT. Let the CIO’s suffer the consequences of their choices.” or “We are seeing a trend in the financial services and telecom sectors where the CIO is actually starting to see the power bill and various users are now getting their individual operations back charged for the cost of power and cooling. There are some pretty sophisticated models being developed internal to many leading organizations in order to calculate and allocate these costs. Previously neither the user nor the CIO seemed to care about the power bill because they never saw it. ” ” So with increasing OPEX from rising energy costs the CIO is challenged with driving IT systems efficiency, which includes energy efficiency but not limited to that concept alone. Shutting down extraneous applications, eliminating software seats/licenses, along with consolidation and virtualization all have a multiplying effect on overall IT system efficiency”
This diversity in opinions come altogether in the statement: “I think there is amazingly a lack of accountability when it comes IT energy efficiency. It’s usually not measured or accounted for so it gets lost among a host of other important tangential issues.”
Some opportunistic behaviour is also reported, the sense of urgency “is highly dependent on the economic state of the company at any given time.” It depends “when times are good, no one cares about power efficiency as they figure whatever money is spent to build a DC out can be justified as contributing a positive ROI. However, when times get bad, then everyone has an opinion about how power efficiency is cutting into the bottom line. The funny thing is, no one can define exactly how electricity usage contributes to the final product or how reducing usage would affect the bottom line because no one person (that would care about such things) seems to have access to all the data.”
You could say that energy cost management still has to find its place in organizations …
Resistance to modernization and innovation
Although there is a lot of knowledge about the do’s and don’ts in power usage this doesn’t mean this knowledge is always used,”I see brand new data centres continuing to be designed and built using old inefficient technology.” Segregation of hot and cold air within data centers isn’t a common thing for everybody. One comment was “Last week when I asked the question again there were less than 5 hands go up out of a room of 200 attendees and that is the best result I have seen yet. If we can’t get to the most basic issues how the hell are we going to get people to do anything else. When I bring it up there is always some reason why they can’t do it but the main reason as far as I can tell is that no one has taken it upon themselves to make this happen.”
Apparently, people don’t want to take risks on new ideas, concepts and technology. The rates of adoption for innovations are closely related to the risk appetite at the individual and organizational level. “IT equipment, especially servers and network equipment, are turned-on and left on regardless of their work load. IT managers are worried if not scared to death about the prospects of one or more devices failing upon re-start so they are left on 24×7.” Risk appetite is the amount of risk exposure that the individual, group or organization is willing to accept. For example The costs associated with additional extra energy usage as a result of redundancy will be taken for granted. The key question is how to ensure that the redundancy is achieved in the most energy-efficient way possible. For a data center it is unlikely that people will be willing to pursue an innovation if there is any uncertainty about the system reliability and/or timely response by well-trained staff or consultants.
For a successful data center, the usage of risk management to make balanced decisions about reliability and energy efficiency is a key success factor.
To be continued
And for those that have an account on LinkedIn there is a poll about:
What is your opinion: Who cares most about enterprise IT energy efficiency?
The pol is still open, so take the poll at: