This vendor-written tech primer has been edited by Network World to eliminate product promotion, but readers should note it will likely favor the submitter’s approach.
The amount of data that is being generated on a daily basis is growing rapidly, placing more and more demand on data centers. Not only do we have connected users actively engaged in generating and storing large amounts of content, machines such as autonomous cars and connected planes generate greater amounts of content by orders of magnitude (Figure 1).
As there is value in almost any data, it is rarely, if ever, deleted. This leads to increasing demands on storage capacity. The alternatives are hard disk drives (HDDs) and solid state drives (SSDs). HDD capacities have been increasing steadily over time, and have been the mainstream form of storage in the enterprise until now. But last year the capacity of HDDs was surpassed for the first time by SSDs – and because SSDs are scaling at a faster rate than HDDs, we will never look back.
SSDs use NAND memory, which has an amazing ability to scale. NAND memory is comprised of storage cells formed on semiconductor material. Improved density was achieved via process geometry shrinks at the die level. However, this method of gaining greater density was nearing its physical limitations based on how closely the memory cells were being squeezed together. Undeterred, the industry introduced a breakthrough in the past three years with the introduction of 3D or vertical NAND. Instead of attempting to squeeze memory cells ever closer together, 3D NAND stacks them vertically on top of each other. This allows SSDs to continue to aggressively scale in capacity for the foreseeable future. Another inflection point took place during the past year – enterprise SSDs became less expensive than 15K HDDs when taking data reduction technologies such as compression and deduplication into account. Compression reduces bits and hence the amount of storage needed for a set amount of data by identifying and eliminating statistical redundancy. This is possible because most real-world data exhibits statistical redundancy. For example, an image may have areas of color that do not change over several pixels. So instead of coding “red pixel, red pixel ...,” the data may be encoded as “100 red pixels.” Deduplication is a special form of data compression that eliminates duplicate copies of repeating data. While compression takes care of repeated substrings inside individual files, deduplication inspects volumes of data to identify large sections (such as entire files) that are identical and stores only a single copy. Storage vendors claim they can achieve approximately a 3x data reduction using SSDs without negatively impacting system performance. The same cannot be said for HDDs due to their inherently slower nature. Therefore, when comparing SSDs and HDDs in a storage array, the relative cost for SSDs is divided by three when taking data reduction technologies into account. SSDs have become more cost-effective than 15K performance HDDs on a per gigabyte basis (Figure 3). This cross-over to SSDs, compared to 10K HDDs, will happen very soon as well. This has triggered a massive transition to flash in the enterprise. SSDs have always had a good value proposition versus HDDs – faster, more reliable and lower power. Adding lower cost and higher density makes the SSD alternative all the more compelling.
Leading vendors recognize the need to offer the densest storage arrays possible and have aggressively designed products that incorporate the latest, high capacity SSDs. At a system level, the benefits of dense SSDs are amplified. For example, the largest HDD available is 10TB in a 3.5” form factor. A maximum of 12 of these drives could be installed in a standard 2U server, providing 120TB capacity. Alternatively, the same server could be equipped with 24 2.5” 16TB SSDs, providing 384TB capacity – or 3x the density. System equipment cost aside, the resources needed to build and operate a data center are considerable. To get a good sense of their significance, consider total cost of ownership (TCO) calculators that estimate the capital and operational expenditures for a given data center. As an example, a modest, tier 3 data center (which supports 99.982% availability) with 10 cabinets of system equipment using SSDs would cost approximately $887,000 of construction and operational costs over three years. The same amount of HDD storage would require a 30-cabinet system (SSDs having 3x the density), resulting in estimated construction and operational costs of $1,423,000. The density benefit alone results in over half a million dollars in savings over 3 years.
It is clear that storage requirements are growing exponentially, driven by inexhaustible content generators that include human and non-human sources. Great value is being derived from this information stream through data analytics, which requires much of the information to be stored on a faster medium than archival tape. This faster medium is transitioning from HDDs to SSDs. Not only are SSDs higher performing and more reliable, they are now also higher capacity and lower cost than performance HDDs. The density advantage translates to even lower costs when building out and operating data centers that host the storage. Businesses looking for a distinct competitive advantage should waste no time in taking note of this trend.