Today’s data center is going through a constant state of flux in an attempt to keep up with current demands. The data landscape grows exponentially, and with that growth comes the need to expand current storage and data center infrastructures. This expansion is a fact businesses in every vertical have come to accept, but it comes with a price.
The Data Landscape
Four billion dollars is spent every year on data center energy consumption and this number will only continue to climb. The type of data growth is also a contributing cost factor; mission critical data is growing in the enterprise environment. This means companies are buying more expensive energy hungry equipment to provide needed fast access and redundancy at both the server and storage level.
Businesses may have accepted that they will have to make necessary accommodations for data growth, but what some have not considered is that this growth has limits. Power is not an infinite resource; in fact, industry experts predict that 96 percent of data centers will not have enough power by 2011.
In a landscape this progressive and complex it is easy to see how a movement like green IT can run rampant. Consumers are bombarded from every angle with green claims. Some green products legitimately lessen data center impact and others greenwash less than environmentally friendly technologies.
And until recently, guidelines to navigate truly green solutions were sparse. But thanks to the U.S. Environmental Protection Agency (EPA), Energy Star standards have been established for Data Center server products. The EPA, along with the non-profit Storage Networking Industry Association (SNIA) and manufacturers in the storage community, is in the process of developing an Energy Star standard for enterprise storage as well.
Measurement Standards
Both SNIA and the EPA realize that there is no silver bullet for reaching green data center status. No one data center improvement will yield a green data center, a comprehensive approach is required. As an extension of that reasoning overall data center greenness should be an inclusive statement.
The EPA uses the metric of power usage effectiveness (PUE). This is measured as the total power entering the data center over the power used by the sum of the IT equipment inside. Numbers nearing one are considered a good ratio, numbers up over 2.5 (the average data center) could use improvement. One estimate suggests that if all current best practices are in place, a PUE of 1.6 can easily be reached.
SNIA measures efficiency as a function of power consumed on an individual device basis using metrics such as Gbs/W, and IOPs/W. Keep in mind, SNIA also takes into consideration the market sector each storage environment exists within. A SAN in a SMB environment would not be judged equally to a SAN in an enterprise environment.
• Gbs/W: A measurement of the number of gigabits per second that are achieved for every watt of power. This metric quantifies physical capacity versus energy consumption.
• IOPs/W: A measure of the Input and/or Outputs per second from disk for every watt of power. This metric shows you the amount of computation that can be delivered for every watt of power consumed. It is the metric that places value on performance versus energy consumption.
These measurements provide a construct for determining overall energy efficiency while placing a value on performance. Additional software functionality is often useful in improving storage efficiency, but effectiveness is dependent on how the customer implements and uses the functionality.
What Does the Green Movement Mean for Data Centers?
The EPA and SNIA have both recognized that the IT community should not have to sacrifice performance and capacity capabilities in the name of going green. Instead they promote solutions that uniquely address data center pain points, while reducing environmental impact. This has been best achieved so far primarily through energy savings initiatives.
What does all this mean for data centers? It may mean that more regulations are coming, and data centers may be held to a standard of environmental responsibility in the future. For the time being however, this means that the IT community is receiving an invaluable set of best practices that will bring to light solutions that cost less for power and cooling, and meet or exceed expectations in performance and capacity. These best practices are a step towards doing more with less, which has been and always will be a key driver of innovation in the IT world.
What Does the Green Movement Mean for Vendors?
The Green Movement has created a huge opportunity for vendors to participate in a competitive environment that has the potential to yield groundbreaking advances in data center technology. The storage industry in particular is running with this concept and continues to introduce technology that is changing the face of the industry.
Unfortunately, they have also dressed up some pigs in lipstick and presented it to the market as a green solution. To try and alleviate this problem, some truly green and compelling approaches are detailed below.
Six Storage Guidelines for a Greener Data Center
1) Control Your Redundant Data. Use block or bit level deduplication to cut down on redundant data. File level deduplication is less effective at improving storage utilization. Delta differencing is an efficient means of reducing redundant data when employing back-up operations. Delta differencing provides block level backup that only saves new changes since the last backup instead of creating a whole new file.
2) Change Your (Corporate) Behavior. Implementing simple changes in corporate behavior such as utilizing data compression can help control increasing data stores. Understand, however, that compression can have an impact on storage performance. One must sacrifice performance to gain physical storage efficiency.
3) Put Your Data Where it Belongs. Many data centers have moved to a tiered architecture. For those who have, it’s no secret that some disks are more expensive than others. Because of the variability in price, different types of storage are used for different types of data. Ideally, the tier of storage should correlate with performance requirements specified by customers in order to meet their SLA needs. While Solid State Disk (SSD) provides tremendous performance gains, it comes with a hefty price tag. Moving highly accessed business critical data to Solid State Disk (SSD) increases efficiency. Other data should be allocated to primary (spinning disk) storage whenever possible. Distribute all other data to lower cost tiers according to their access patterns. With multiple tiers of storage, you gain efficiency, however you also increase management costs. Look for solutions that have automated management software included, to help you understand how much SSD you need to optimize your application. Automated management is useful in overcoming the complexity of data management, allowing the customer to view a pool of storage, and allowing the disk subsystem to logically optimize the utilization of storage.
4) Virtualize Your Storage. Rather than having pools of storage directly assigned to individual servers, consolidate your storage into a SAN, and allow the storage devices to allocate space to a wide variety of servers. Virtualized servers allow for more efficient server utilization. Storage virtualization consolidates resource to avoid ‘islands of storage’. Allowing storage to reside on a server needlessly is wasteful when other servers do not have the necessary storage resources to optimize applications. The name of the game is eliminating wasted/underutilized capacity. Storage virtualization better distributes storage between servers for a more effective approach. Storage virtualization also opens the door for technologies such as thin-provisioning which are used to improve utilization, but can result in management overhead.
5) The Virtualized Daily Double. Going to a fully virtualized data center carries huge green potential as far as reducing the amount of total hardware. Virtualized servers and virtualized storage taken collectively represent a power play in the data center. Server virtualization has a huge impact on the number of physical servers needed to run various applications, and virtualized storage reduces hardware requirements and enhances capacity management capabilities.
6) Replace Legacy Equipment with Energy Efficient Alternatives. The greenest implementation possible is to run all the preceding guidelines on energy efficient storage platforms. This can be advantageous from a performance and capacity play as well. Extremely dense storage solutions exist that provide high capacity, while the sheer numbers of spindles provide unprecedented performance gains. Because of their compact efficient design these solutions can provide a 50-80% power and cooling reduction with similar performance and capacity compared to legacy systems. It also occupies less real estate, which means storing more data in significantly less space.
Steve Visconti is the president of Atrato, a data center solutions provider.
Photos CC-licensed by Flickr users Random Activity and ChrisDag.