Last week’s Stakeholder meeting for Energy Star for Storage was very well attended and an excellent example of diplomacy and tact in action. I was so impressed with the way that all the parties with diverging interests made their point patiently as the complexities of the emerging technologies in storage evolved and created obstacles to simple language and easy to use guidelines.
The most important and urgent reason for convening this stakeholder meeting is that in the case of servers, assets were sitting idle and went unnoticed until experts began to assess their load. This might well be true in the case of storage devices as well, and as an industry, we do not yet know whether the ROI on those assets is being well used, or wasting away. So, with the premise that with well defined states, you can collect data, it is time to focus now on the topic of storage energy efficiency.
Representatives from Climatesavers, Ecos, EMC, EPRI, Fujitsu, Hewlett Packard, Hitachi, IBM, Intel, NetApp and Oracle filled the audience in downtown San Jose at this meeting facilitated by Stephen Pantano and Al Thomason of ICF as well as and hosted by Una Song, Energy Star Lead, Storage for the US EPA (Environmental Protection Agency)
At the highest level, the disk and controller configuration are the most meaningful factors for energy. These need to be defined by capacity not access. Should these be defined by controller quantity and controller configuration, followed by disk quantity and disk configuration?
The challenges understood at the beginning of the meeting are:
What can be qualified?
How can it be sold?
How will it be used?
What will customers fix/replace, improve after their initial purchase that might make the systems more or less energy efficient?
First there were definitions to sort through: ready idle versus deep idle, offline idle v online idle, resulting in discussions about the intent v the naming of a particular point in time within a technology. That was followed by a ‘data management and output’ discussion that eventually gained agreement that power and temperature fluctuations are not as well defined in storage as on servers, and latency, not frequency is the most important parameters to measure. It was also observed that the quantification of energy efficiency savings will be difficult on some key elements of storage, for example on de-duplication. So this is one element that might be noted and set aside for future data sets or future versions of the Energy Star specifications.
Some small samples of storage appliances yielded preliminary data that there is little difference between power supply usage, whether they are on servers or storage, but this needs to be validated with a larger study.So the proposed scope of the first version of the Energy Star guidelines will include online and removable storage, and then expand to larger systems and near-online products in specific configurations. The plentitude of customized configurations is the conundrum in this committee – there are so many ways to design storage depending on technologies, pricing models, space, cooling, etc. that conforming to specific configurations is not a viable outcome.
The EPA/ICF team was very pleased that this meeting had elicited comment on a lot of the complexities of the potential guidelines and said they didn’t mind at all that no clear decisions had been taken – “too early in the process”, said Stephen Pantano. Have a green day!