Storage Peer Incite: Notes from Wikibon’s October 30, 2007 Research Meeting
Moderator: Peter Burris & Analyst: David Floyer
This week Wikibon presents Storage: The heat is on. Data centers consume 1.5% of all the power used in the United States, according to a recent U.S. Environmental Protection Agency report, and this ratio is probably true across the industrialized world. This is an increasingly expensive statistic on corporate bottom lines as the cost of energy increases and the need to reduce carbon footprints becomes more urgent.
Storage is an important part of this, particularly as demand rises. But the answers are not always the same for storage as for the compute side of the IT shop. In IT every discussion of energy reduction starts with one word -- virtualization. But in storage while this can reduce the number of disks spinning, it increases controller complexity, so that on average virtualization only reduces storage energy use by 5-10%. For greater energy use cuts, IT must look to other strategies, including choosing vendors and products for their energy efficiency, moving to modular storage where that is practical, and developing efficient tiered solutions that maximize use of tape, still by far the most energy efficient method of storing data and useful for tier 3 where 60% of stored data resides. Bert Latamore
Contents |
Storage: The heat is on
Energy concerns are gripping not only data centers but increasingly the corner offices of large organizations as oil and other energy prices push north of $90-$100/barrel. Large storage suppliers are investing in energy efficiency-based product differentiation. These concerns are real for data centers. A recent EPA study claims that 1.5% of all energy draw in the United States ultimately is consumed by data centers and that without significant adjustments in technology and consumption behavior this figure is likely to increase.
Users face multiple challenges as they try to improve the energy profile of their storage. Storage today accounts for approximately 10%-20% of total energy costs in a data center. More than 50% of energy costs are consumed by the infrastructure that supports equipment -- including cooling, power distribution, and UPS technologies -- all of which vary by IT equipment. Consequently the effect of investments in design, product selection and implementation of more power-efficient storage products will dramatically impact the power consumption of the data center's infrastructure.
We see five action items effecting near-term user positions:
- Users and vendors must recognize that the largest contributor to data center energy costs remains the overall design, structure, architecture and location of data centers themselves. Data center environments remain the 800 lb gorilla in all IT energy conversations. Users must plan to use any technology that can improve environments, including liquid cooling, that can reduce that energy use, and the Wikibon community is encouraged that most data center managers are anticipating a return of liquid cooling and other actions.
- Users should explore modular storage technologies where they provide an overall better energy footprint than traditional arrays. But any move to take advantage of modular systems must be combined with efforts to better associate applications with performance and capacity requirements as storage decisions are made.
- Tape is a green technology and must be a feature of any organization's Tier 3 strategy for it to truly claim to be a green shop.
- Users must determine the differential they are willing to pay for green technology and begin expressing their overall green objectives to the industry not only through purchasing but also in their communications about the storage products and product mix they are likely to make on future purchases. We see multiple types of products that will lead to greener storage in the pipeline worldwide. However, these products won't see the light of day soon unless users consistently demonstrate that green is a central feature of ongoing strategies.
- Finally, none of this will accomplish anything unless users constantly demonstrate that green values are real and not something to be negated at purchase time in favor of lower cost, less energy-inefficient alternatives.
Action Item: Green storage, like other green initiatives, is more than good citizenship. It will lead to overall operating cost savings in the long run. Near term, IT organizations need to force attention on overall data center design and establish long-term plans for the adoption of green-related technologies, pushing the industry to improve overall product efficiency. Additionally some large vendors of storage technologies have developed competencies for comparing as is and to be implementation options. They should be pressed to provide knowledge of the energy use implications of different configuration alternatives in terms of their overall impact on the greening of the data center, including the impact on the infrastructure that uses more than half of the energy data centers consume.
Green storage is all about the other green
While good corporate citizenship is marketable, green storage initiatives are being driven primarily by one thing: a reduction in operating expenses. As one IT executive at a major U.S. bank put it: I’m going to be totally honest with you. I want to save the planet. I have children. I want them to breathe clean air but trust me, in corporate America, and even corporate Europe, people aren’t going to do this if they’re not going to save any money.
Storage environmentals can comprise as much as 20% of total operational costs. Operational cost reductions are measurable, and tying IT cost reductions to an increase in share price is simply good business for CIO's. The question is: How far do corporations want to go and what are they willing to do to get there?
Organizations have a choice: ride the wave of incremental innovation today's storage products are projected to deliver for reducing cooling and power consumption or send a message to vendors that they're willing to pay more now for dramatically lower operational costs in the long run. While everyone is talking about green storage, this message has not been put out and until it is, earth-shattering cost reductions should not be expected.
Ultimately, lines-of-business will have to drive green solutions because they're dramatically cheaper in the long run. But if this is not perceived as the case within organizations, technologies that deliver dramatic reductions in energy consumption may never see the light of day given that vendors must generate an economic return. The CEO, CFO, CIO and Board of Directors must have this discussion and determine how hard they want to push their organizations and suppliers by investing in less mature, more expensive (to acquire) green technologies.
Action Item: Tactically, it is imperative that IT forces business lines to fully factor energy costs in procurement decisions. IT must develop metrics that facilitate that process and verify the viability of green initiatives by determining the degree to which the organization is willing to pay up for greener technologies. A demonstration of real commitment to this issue will come in the form of internal tax credits and unambiguous incentives for initiatives dubbed as green.
MAID - Massive (Misunderstood) Arrays of Inactive Disks
The MAID concept was developed in 2001- 2002 and brings the acquisition price and operating expense of disk storage closer to that of an automated tape library. MAID was the first effective "Green Disk Array" to appear. Five years later, many issues about MAID are not well understood.
For example, with MAID, disks are powered off when not in use yielding considerable energy savings. But powering disks up and down becomes a problem if they are FC-, SAS- or SCSI-based. These disks are made to be powered up and run 7x24 for a few years without spinning down. However, the popular desktop SATA drives are designed to be powered up and down. Manufacturers ratings and lab tests confirm ~ 30,000 power up/down operations can be expected from SATA devices.
To date this point has not been effectively communicated by the industry and misperceptions persist. With a MAID array, no more than 25% of all drives are powered up at any given time, significantly lowering power consumption while prolonging the life (MTBF) of the disk drives. The acquisition cost of MAID subsystems is lowered as the MAID controller has to support only 25% of the memory, I/O connections, and total devices at one time. Today MAID is best suited for WORSE (Write-Once-Read-Seldom-if-Ever) applications and not high-performance tier 1 workloads.
Action Item: Misunderstandings about MAID have slowed its market impact to date. MAID suppliers need to develop a compelling marketing message and crisper positioning to realize the full impact and benefits of this potentially beneficial technology.
The environmental balance sheet
IT has the potential to save energy, but what the business is (or soon will be) asking is why IT is using so much power, and what can be done about it. IT should respond by giving a clear mandate to a “Green Czar” to:
- Measure the power IT uses now, for storage and other technologies,
- Project future power use,
- Establish the energy saving contribution of the applications portfolio and projected applications,
- Create a balance sheet of energy usage and energy savings and project forward,
- Establish the initiatives that will impact energy use and agree with the business and IT on what actions should be taken,
- Integrate the information into a report that can be used by the CIO and the business to establish the organization's IT green credentials.
Action item: Set a senior executive the objective of producing an energy balance sheet early in 2008.
Planning for higher storage power densities
The important metrics for storage equipment are power consumption (kVA) /heat production(kWatts) and power loading (kVA/sq.ft.)/heat loading (kWatts/sq.ft.). The kVA and kWatt figures will allow the cost of power and cooling to be calculated. The kVA/sq.ft. and kWatts/sq.ft. figures will establish if sufficient power and cooling is available in the data center. The important trend for storage is that the power and heat densities are going up. Controller technology is getting denser, and the move from 3.5” to 2.5” disk technologies will increase drive power densities.
The current power density for high-end storage is about 0.7kVA/sq.ft. Wikibon projects predicts that this figure will increase to over 2kVA/sq.ft. over the next five years. This figure is not as high as projections of over 10kVA/sq.ft for racks of 1µ server blades, but most data centers have power densities of significantly less that 1kVA/sq.ft at present.
Action item: Storage vendors (perhaps under the leadership of organizations like SNIA) should give clear guidance on future power and heat metrics. IT departments need to plan now for significantly high power and cooling densities in the data center.
Will high energy costs mean storage technology substitutes?
According to PricewaterhouseCooper's most recent MoneyTree report, interest in green technology is smoking hot, up 80% sequentially in the third quarter of 2007, capturing more than $840M in venture investment for the period. A clean tech company (GreatPoint Energy of Cambridge, MA which converts coal and biomass into natural gas)inked one of only two deals worth $100M or more in the quarter. The MoneyTree report also indicates that information technology remains the sector capturing the most investment, attracting $3.7B in the quarter.
What does this have to do with storage? It seems plausible that among portfolio companies of VC's, the intersection of clean tech and IT will lead to investments in data center technologies that are greener, including those within the storage infrastructure domain. While complementary technologies such as storage virtualization, thin provisioning and data deduplication --which are being retrofitted into existing product sets-- will get a lot of attention, alternative storage technologies may very well reemerge in this equation.
Specifically, two technologies in the storage heirarchy bear watching, tape and solid state disk. Tape is the greenest of all storage technologies. By its continued existence and persistence, it appears an excellent candidate for innovation, which has been lacking in recent years. Tape remains a multi-billion-dollar market, and, with most of the world's information residing on tier 3 storage, tape, both as a standalone technology and in the form of hybrid disk/tape approaches, could attract investment. Solid state disk (SSD) is another green(er) technology that could make a comeback. As disk aerial density improvements slowly flatten, semiconductor-based alternatives could become more attractive and challenge disk for TB consumption as green storage heats up.
Other areas that bear watching include storage services, distributed file systems and liquid cooling of storage circuitry. Storage services are a natural for reducing power consumption as outsourcing storage will immediately lower the power bill. Perhaps as interesting is distributed file systems like the Google File System that spread data everywhere, presume component failure and run on mega-data-center infrastructure. This type of approach strips out the need for complex controllers and boils the power problem down to the disk device itself, which can then be attacked specifically. Finally liquid cooling of storage control circuitry will come indirectly from processor innovation, not likely from storage suppliers. However it will seep into the marketplace, albeit at a premium price.
Action Item: New invention and efforts to build companies will probably coalesce around green storage. Vendors wanting to participate and not get left out in the cold must shift the emphasis of green from marketing to more substantive research, invention and development.
Action Item:
Footnotes: