Read the Final Wikibon Submission to the EPA's initial ENERGY STAR framework.
On June 4 the Environmental Protection Agency (EPA) issued an Energy Star Enterprise Storage Draft Specification (footnote 1) with a cover memo (footnote 2).
On June 26 3PAR shared with Wikibon its feedback to these proposals. Generally, Wikibon agreed with 3PAR's assessment and solicited feedback from its community which was quite useful in formulating our final submission. We'd like to thank all those who contributed on this post, through email backchannels, via Twitter and the good old phone system.
The feedback from Wikibon members emphasized:
- A general agreement that explicitly including software in early discussions would be useful;
- However, several members expressed concerns about the ability to demonstrate the efficacy of software innovations in a consistent manner and in turn having the EPA essentially 'mandate' the inclusion of such innovations;
- A general concern over the conflict between the need for enterprise arrays to accommodate high availability and the potential negative impact redundancy might have on perceived energy efficiency;
- A clear sentiment in the Wikibon community that we should positively support the EPA in this endeavor and not come down too hard on possible omissions, especially at this early stage.
- A continued belief that unplugging disks is the best way to reduce the energy bill-- i.e. classifying data and getting rid of stuff is still the best method to reduce costs, albeit an elusive one.
Generally Wikibon continues to believe software is a key enabler to reduced energy consumption in disk arrays, especially in the context of an Energy Star rating. This opinion is based on the simple fact that the largest component of power consumed in arrays is from spinning disks; and techniques to eliminate or minimize disk drive power consumption will have the biggest impact on improving energy efficiency.
As such we are asking the EPA to allocate time to discussing the accommodation of software innovations on the agenda at its July meeting in California. While we understand there are many issues related to establishing baselines, field validation, lack of standards, etc, we strongly feel the topic should be openly addressed. We are concerned based on our reading of the draft document issued that software impacts are being buried without due vetting.
We are not suggesting that power supplies should be ignored as part of the EPA Energy Star specification for storage. On the contrary, we believe that incentives to improve power supply efficiency are warranted. As well, we feel techniques can be developed to rationalize the apparent conflict between high availability and redundancy (e.g. N+1 approaches).
However, we feel our initial analysis of this issue warrants consideration, specifically:
The EPA's strategy to pressure manufacturers to improve power supply efficiency has worked well for PC's and laptops, where single power supplies provide sufficient reliability. As such, the EPA was able to motivate PC manufacturers to pressure the power supply manufacturers to increase their efficiency which is of course beneficial to the market and environment as a whole.
Enterprise disk storage is a completely different matter. If data "in flight" between a PC processor and a PC disk is lost due to a power supply failure, it is "inconvenient" but not catastrophic. In the enterprise storage space, with the exception of the low end JBOD market segment, customers require redundant power supplies and internal batteries to guard against "in-flight" data loss, which could lead to catastrophic disruptions to mission critical IT services.
In the proposal the EPA considers the adoption of a 'Net Power Loss'. Under this approach, EPA would either specify the allowable absolute power loss (in watts), or specify the minimum efficiency for a power supply under actual use conditions(i.e. efficiency at idle and maximum power load of the storage system). Wikibon is concerned that this Net Power Loss approach would mean that storage systems with redundant power supplies would be unlikely to comply with the EPA standards. The only storage arrays that would be likely to comply would be entry-level "JBOB" ("just a bunch of disks") systems with single power supplies.
The key point that Wikibon would like to emphasize is that the potential savings in power with the improvement of power supplies is limited to about 10-15% of a subset of the storage array market.
Far larger savings can be realized with software functionality in storage arrays. Software allows two main benefits to be achieved:
- Storage array software can reduce the power or turn off disk drives, which account for 80% of the power of a storage array.
- Storage array software can reduce the amount of data that is stored on the disk. Techniques for reducing data on the disk include:
- Thin Provisioning
- Deleted data reclamation
- Tiered storage
- Flash, which like magnetic disks storage retains data when power is turned off, uses much less power and can be turned on and off very fast.
These techniques are in the main additive. The potential for saving data using one or more of these techniques is again to achieve an additional 80% or greater reduction in the amount of data stored.
In the enterprise storage space, the savings from improved power supplies could achieve 10% savings. The savings from software could achieve more than 50% savings. The savings from software dwarf the potential savings from power supplies.
Wikibon has been working with PG&E and other utilities in helping users receive incentive payments for storage software solutions. We are confident that a software approach could be defined.
Wikibon does not believe that the current EPA proposals will advance the development of green storage and green data centers. Many storage products will not meet the power supply requirements, but will offer far higher levels of savings with other software-based technologies. This specification will drive the sellers of these products to discredit the Energy Star brand, especially in competitive situations. The results are likely to confuse buyers of storage technology and a reduce the effectiveness of the EPA and Energy Star affect meaningful change in energy behavior.
Action Item: EPA should include software functionality in its specification for achieving energy star. This would allow a far more aggressive energy savings to be set as a standard for Energy Star certification. The vendor should be given the choice of how to achieve these energy savings against the base of a storage array with no software and poor power supplies. This approach will achieve higher levels of savings and enhance the EPA energy star brand.