It is important to note that one of the sides of the storage efficiency triangle is not deduplication or compression. Technologies like deduplication and compression have their place, especially in backup and archive applications. While there are some corner case exceptions, capacity optimization technologies do not have near the return on the investment in production storage that they do in the archive and backup use case.


The first side of the triangle is energy efficiency. Selecting systems in disk storage that use less power is now a critical consideration. This increase in importance isn’t the result of a newfound environmental concern nor is it really being driven by reducing the electrical bill. While both are definite benefits of an energy efficient system, the primary reason for its rise in importance is simply that many data centers, especially in major metropolitan areas, can no longer get more power. Due to a number of factors, like an aging power grid and the difficulty building power plants, many data centers can’t buy additional power, at any price - or at least it involves a major effort.


As a result, even disk storage systems need to be more power efficient. A good place to start is with the disk drive itself and MAID technology. MAID (or Massive Array of Idle Disks) allows drives to be spun down or even turned off when not in use. Traditionally, this technology was deployed for archive or backup scenarios. Using methods like Nexsan’s AutoMAID technology enables this power management technology to be used effectively on disk storage systems for a wide range of applications including primary applications.


In the past MAID technologies have involved too much latency, or the time needed to return the drives back to full speed, to be used in the primary storage tier. This latency could cause users to wait for disk response and applications to time out, in some cases. To be valid for primary storage MAID technology needed more sophisticated power savings modes in order to reduce latency. These newer MAID technologies establish more granular power savings levels or modes, each producing greater power savings, with a corresponding amount of latency. As data becomes more idle, a longer spin up time can typically be tolerated.


Another important aspect is that these power efficiency processes need to happen automatically. Storage managers are busy enough and don’t need or have the time to analyze all their data to select which goes on each MAID level. Typically,they know at a high level which volumes are very active, like database or email, and which volumes are less active, like user home directories. With modern MAID technologies, if the storage manager simply segregates at the macro level, which they likely have already done, MAID storage can set different power-down parameters for each volume, extracting power savings across the environment.


The second side of the storage efficiency triangle is space efficiency, or using high-density storage. Again, often thought of as a more appropriate in backup or archive applications there is a similar requirement for density in primary storage. In the same way that data centers did not suddenly become environmentally conscious they also did not become space-conscious. As is the case with power, there simply is not enough physical floor space to house all the storage that the modern data center requires.


Again, a quick answer here may be to leverage thin provisioning, deduplication or compression technologies. The challenge especially in disk storage is that these technologies find limited over provisioned volumes, duplicate data or compressible data. While all of these technologies can provide value they should be layered into the storage system after it is efficient in its own right.


The more appropriate solution may be to use high-density shelves with high-density drives. Nexsan, for example, can support 42 drives in a 4U chassis thanks to its top loading design. This can deliver up to 84TBs of storage per system, replacing what previously might have taken multiple racks to house and power.


Obviously, there are reliability concerns in systems that are this dense and utilize drives of this capacity. High-density systems must take significant measures to make sure that data is not lost and that the system can support the performance capabilities of that many drives. The system should have dual controllers for reliability and those controllers should have multiple RAID engines to be able to sustain the performance required. Additionally, the system should perform continuous background data integrity scans to ensure drive health and readability when data’s accessed.


How companies achieve this density of storage is important. Simply packing more drives into a heavily customized drive pack can lead to cost and availability issues, as an entire pack must be replaced when a single drive fails. Density should not come at the expense of a high TCO. Ideally, the system should have individually replaceable components allowing just the failing drive to be replaced instead of multi-drive blades.


Finally, the system should have a built-in management interface that monitors and alerts the storage managers to the status of the system. For many environments, one or two of these 4U systems is all that’s needed. While this does make the storage manager more efficient because there are now less systems to manage, it also means they have fewer systems to cover a failure. Failure must be identified and quickly fixed. It is important that the status of the system be known at all times and that critical alerts be emailed to a team of individuals, allowing problems to be addressed well before they affect the application or user’s experience.


The final side of the triangle is cost efficiency. The first two sides, power efficiency and space efficiency, help in this respect. Power efficiency can reduce the electricity and cooling lines of the budget. Space efficiency can reduce the cost of acquiring additional data center floor space, as well as making the storage managers more efficient, with  fewer systems to manage. Also, with space efficiency there is less physical packaging to pay for. With the class 12 or 15-drive, front mounted drive shelf, each empty bay has a rack-space cost. Space efficient systems reduce that.


Costs can also be controlled by getting maximum performance and reliability out of less expensive, higher capacity drives. All drives take roughly the same amount of space, regardless of capacity, so utilizing the technology to its fullest and switching to larger capacity drives can greatly reduce cost.


At the end of the day, the simplest and most direct way to reduce storage costs is to make the systems themselves are less expensive. This requires that the storage system manufacture itself be efficient by producing reliable systems that are easy to use and maintain. Doing so saves them from having to build out huge and expensive sales and services organizations that need to justify the cost of the system. In short it really shouldn’t take an ROI spreadsheet to show you that the storage is less expensive, it just should be less expensive.


Disk storage is often the greatest drain on the storage budget. Making this storage tier as efficient as the downstream tiers of storage can have a great positive impact on that budget. To achieve disk storage efficiency systems need to better deliver on the three sides of the storage efficiency triangle regardless of the tier of storage they address.

George Crump, Senior Analyst

This Article Sponsored by Nexsan