SAN/NAS

By Randy Chalfant

Growing demand for capacity in ever-shrinking data center space has motivated some storage vendors to cram as many disks as possible into a single space to manage capacity needs without expanding the storage footprint. As much as some vendors call the problem solved, it’s not that simple. Without proper efficiencies and architecture, high density storage systems can create more problems than they solve and ultimately devastate your business. For some, that comes in the form of reliability issues with poorly constructed storage subsystems. For others, it comes in the form of the power/density paradox that hinders end users from utilizing the very floor space that high density systems were intended to conserve.

Let’s examine. The total amount of power an equipment rack uses has grown with dense packaging, i.e., higher density = higher kWs per rack. Higher power densities per rack are posing substantial difficulties for data centers that were built at a time when 2-3kW per rack were common and more than enough power existed to support them. However, the average power per rack has doubled to 6-8kWs and as high as 15kWs for very dense packaging. The issue for data centers compounds with inefficiencies in power distribution and the need for additional cooling. Buildings with infrastructures designed 10 years ago are now reaching their limits on what they can support as data continues to increase at exponential rates.
To illustrate the point, the EPA stated in a report to Congress that many data centers have already reached their power capacity limit:

•    More than 10% of data centers (large medium and small) are already out of power; 
•    68% expect to be at their limit within the next three years;

The EPA report also states that power and cooling for storage represents 40% of total IT expense; and energy costs will increase by 19% CAGR from 2006 to 2011. In a weak economy, data centers are increasingly motivated to extend the useful life of everything from technology to facilities. As such, vendors have responded with ever increasing amounts of capacity packed into a given amount of space

What are the risks of high-density storage? What is the power density paradox? How does can a storage vendor help customers address both? This brief explores these important business and technology issues.

The Power Density Paradox Illustrated
With the costs associated with new data center construction, the allure of a high-density rack is attractive. However, without improvements to subsystem power efficiency, high-density storage leads to a problem known as the power/density paradox, as the following scenario illustrates.

Imagine a 4U storage subsystem with 60 drives. Without advanced power management to reduce the power consumption of a storage array, one could expect an average power draw of 1,750 watts with a peak load of 3,904 watts. A 4U subsystem would allow 10 subsystems in a rack which would produce a power draw of over 15.3kWs per rack. That means only 30 racks could be put into a 10,000 square foot facility! By trying to solve a space problem, high density solutions end up creating an entirely new problem - power! As most facilities don’t have enough available power to expand, organizations are left with few alternatives as they walk through a now spacious datacenter that is out of power.

In a storage environment, each rack requires a certain amount of space for aisle room as well as heat dissipation which must be removed from the data center through air conditioning. Ten years ago, a rack that may have drawn a total of 1kW would have about 65% of the available floor space dedicated to aisles, cooling, UPS etc. That left 35% of the floor space available for racks. Based on those numbers, a 10,000 square foot data center could have supported up to 500 racks on the data center floor. In a data center today, by increasing the power drawn to 10kWs, only 50 racks could be housed in 10,000 square feet .

Many data centers end up trading space for power resulting in more empty space without any real value achieved, thus, the paradox.

The Power/Density Paradox Compounded: The Cascade Effect
It may come as a surprise to find out that for every one watt of load added at the component level, there is an additional 1.84 watts used through power inefficiency from the time power enters the building until it is used by a device ¹. Likewise, for every 1 watt reduction at the component level (processor, memory, hard disk and so forth), there is an additional 1.84 watt savings on the energy bill. This is known as the cascade effect and is the result of inefficiencies, or loss of power, from power supplies, power distribution systems, UPS systems, cooling systems, building entrance switch gear, medium voltage transformers, etc. So what can IT professionals do to mitigate this paradox?

If the power/density paradox is the problem and the cascade effect compounds the problem, what is the solution? The only solution to the paradox is to decrease the total amount of energy used for a given storage density. That can be done by limiting the number of components within a subsystem or by the subsystem itself being more power efficient. Technologies that uniquely take advantage of the cascade effect to change the outcome of the power/density paradox by automating power management provide a solution to the paradox and a company’s storage requirements because they can significantly reduce the power consumption of a disk array without reducing storage performance - speed with green.

Management of power consumption through a policy system can save significant amounts of power in the data center. By using high density storage systems with automated power management, IT professionals can grow capacity without data center sprawl while reducing operating expenses. For example, because of the cascade effect, when the power load is reduced by one watt, a total of 2.84 watts are saved1. What happens is that every attempt to save energy at the component level ends up becoming a cost reduction multiplier. For every watt saved by a subsystem, organizations get nearly three watts saved at the meter.

Power Saving with newer MAID technology
By automating the power management process, users do not suffer the limitations of first generation MAID technologies that took an “on-off” approach to power efficiency. These new power management or MAID 2.0 approaches provide the benefits of power efficiency without the performance limitations inherit to the MAID 1.0 “on-off” approach. As such, storage systems leveraging MAID 2.0 has the ability to spin down drives to lower energy consumption between data references. Whereas old MAID technology suffered from the inability to quickly recall data when needed, MAID 2.0 delivers sub-second response times to the first I/O request and remains at full speed for every subsequent I/O request until enough idle time has elapsed to activate energy savings once again.

Whereas most might view high density storage as a way to free up facility space, there is a power/density paradox that must be addressed. Policy-based automated power management or MAID 2.0 is the clear approach to that risk. By combining energy savings with high density, IT professionals get what they need most — more capacity while reducing floor space and costs. And with MAID 2.0, those cost savings can finally be achieved without reduction in storage performance. In a new world of data expansion and economic contraction, storage efficiency is king.

Randy Chalfant is vice president of strategy for Nexsan Technologies. His expertise is represented by more than 35 years of experience in business development, storage, mainframes, open system servers, operating systems, applications, and networking solutions.

[1] Source: Emerson Network Power, White Paper: Energy Logic: Calculating and Prioritizing Your Data Center IT Efficiency Actions