Data storage demands continue to increase. The value of information is ever increasing as are the requirements to retain data for longer periods of time. Consequently, the amount of data being generated and stored annually continues to grow in environments of all sizes.
With this relentless growth comes the expense of powering and cooling the IT systems generating, utilizing, and storing data. Environmental and energy topics have joined the list of business and technology related issues that executives and IT managers must consider.
The top three power consumers in the data center are heating, ventilation, and cooling (HVAC); servers, and storage. Reducing the amount of power consumed in the data center means improving efficiency. HVAC power consumption can be reduced with more efficient cooling. Server power consumption can be reduced with more efficient servers as well as by using servers more efficiently through virtualization and consolidation. For storage, it also means increasing efficiency.
The storage components that consume the most power and, as a by-product, generate the most heat are the disk drives. Improving the power consumption and cooling requirements of storage means reducing the number of drives in the storage solution. And the best way to do this, for primary storage, is with efficient drive utilization.
There are several ways to reduce the number of drives in your storage system and to improve utilization. The most obvious is through consolidation. Storage consolidation increases utilization from 50% to around 75% because it allows administrators to draw from a unified storage pool to more effectively allocate storage to servers and applications on an as-needed basis. This reduces the amount of capacity overhead required and eliminates unused storage, which is wasted in a distributed environment. Use of storage consolidation alone can result in a 25% reduction in power/cooling requirements due to more efficient drive utilization.
While storage consolidation is one important aspect of efficient drive utilization, there are many others that can significantly impact the number of drives needed in a storage system and thus impact the system’s power/cooling requirements.
Storage Selection Criteria
Since many primary storage systems use the same disk drives, it may be hard to differentiate one system from another. Not all storage systems have the same power/cooling requirements. The same number of drives in one system vs. another may result in systems with equal power/cooling requirements, but the efficiency of those systems can be significantly different.
Storage is typically purchased with one of two goals in mind — meeting capacity requirements or meeting performance requirements. For secondary storage capacity requirements, the goal is to get the most capacity for the lowest price (while maintaining a base level of reliability). For primary storage performance requirements, the goal is to lower the cost per Terabyte (TB), reduce the storage footprint, and minimize power/cooling requirements while maintaining application performance and service level agreements (SLAs).
Capacity-Oriented Storage Requirements
When purchasing storage to meet secondary storage capacity requirements, IT managers are looking for the lowest cost per terabyte. This is achieved with a combination of low-cost, high capacity drives and efficient capacity utilization.
Capacity-oriented applications are based on usable capacity available to store and access data. Power and cooling requirements are based on raw capacity because all the drives in the storage system must be kept spinning and must be cooled. Capacity used for formatting, RAID protection, configuration metadata, and other overhead, must be subtracted from the raw capacity to determine the usable capacity.
Systems that support the highest usable capacity as a percentage of raw capacity can meet capacity requirements with fewer drives, thus lowering power and cooling requirements.
Performance-Oriented Storage Requirements
While performance-oriented storage is also impacted by raw vs. usable capacity; another important consideration is performance per drive. Just as you want to get the most usable capacity per drive for capacity-oriented requirements, in order to lower power and cooling cost for performance-oriented applications you need to get the most performance out of each drive operations.
Performance for primary external storage systems is characterized by one of two metrics — I/O per second (IOPS) or throughput. IOPS is the key metric for transactional applications, such as OLTP, databases or email, with a small-block, random access application profile. Throughput performance is measured in Megabytes per second (MB/s); and is important for bandwidth-oriented applications, such as video streaming, surveillance and other rich media applications. Meeting IOPS or throughput performance needs with the fewest number of drives results in lower power/cooling requirements.
As disk drive capacities have increased, the cost per TB has decreased dramatically. This has been beneficial for throughput-oriented applications that can typically meet performance requirements with a relatively few number of drives. These capacity-intensive applications benefit from the large-capacity drives because they are able to store more data, for less cost, and still meet performance needs.
Using the same strategy for transactional applications produces the lowest cost per TB, but will probably not deliver adequate application performance to meet SLAs for primary, IOPS-sensitive applications.
Industry Consortiums Addressing Green Issues
There are several industry consortiums looking at standardizing and promoting data center energy and performance efficiencies. The Storage Networking Industry Association (SNIA) has established a Green Storage Initiative (GSI) to promote the education and awareness of power issues related to storage. Under the direction of the GSI, there is also a Green Technical Working Group (TWG) whose charter is to define green storage metrics and methodologies to test to those metrics for reliable, repetitive comparison and improvement of storage systems’ power utilizations and efficiencies.
The Green Grid is dedicated to advancing energy efficiency in data centers and business computing ecosystems and has recently partnered with SNIA to work together on storage-related metrics and improvements.
The Storage Performance Council (SPC) has created benchmarks to address the challenge of determining the “real world” performance that users receive from their applications and will likely add a power utilization/efficiency component in the future.
And while not a consortium, the Environmental Protection Agency (EPA) is expected to create an Energy Star program for storage systems, once its program for servers is completed.
Action Item: With soaring energy prices, IT managers must keep energy consumption in perspective without negatively impacting application performance. Increasing the performance of storage systems is one way to reduce energy costs. Buying storage systems on the basis of cost per capacity alone may create performance and energy issues. When purchasing external storage systems, performance combined with energy efficiency should be considered. Improving efficiency for HVAC, servers, and storage is the key to reducing power consumed in the data center.
How will you factor power requirements into your future storage purchasing decisions?
Footnotes: