>Customer demand for storage solutions is continuing to grow. Cloud data centers can help meet the need, says Collins, vice president of product marketing at HGST, a Western Digital company. Collins offers suggestions on how solution providers can help rebuild their clients' systems to support the types of information they want to store.—Jennifer Bosavage, editor
Opportunities abound for channel partners to differentiate and create new revenue streams as storage takes center stage enabling the cloud data center. Demand for storage is exploding as IT managers strive to handle the avalanche of new data being generated by cloud data centers, big data analytics, social networking, HD video and millions of mobile devices. Solution providers can help customers correctly and efficiently re-architect storage to meet that demand. They can support the type of content they want to store, dramatically lowering total cost of ownership (TCO), while ensuring they become preferred channel partners.
[Related: How to Get Your Arms Around Big Data ]
There are five steps to effectively selling cloud storage:
Step 1: Evangelize the New Storage Equation: Beyond Price to Value
Hyperscale datacenters are designed for tens of thousands of servers handling petabytes even exabytes of data. Cloud architects are developing new best practices that demonstrate the bottom line value of taking a systems-wide approach to storage. These cloud data centers are highly customized. Often based on open-source hardware and software, they deploy drives optimized for specific applications that require efficient power usage, density and reliability. The bottom line: Cloud data center decisions are now based on value rather than pricing and TCO is how value is being measured. Best cost-per-TB, watt-per-TB, TB-per-system weight and TB-per-square foot are the new metrics for managing better costs, performance, scalability and storage efficiency.
Step 2: Focus on the Right Mix of HDDs and SSDs
Contrary to popular belief, HDDs are not giving way to all-SSD cloud environments for the foreseeable future. Datacenters are driving resurgence in demand for HDDs with IDC research projecting 46 percent compounded annual growth in this segment. (CAGR: 2011-2016). Although SSDs have many benefits (including ruggedness, better small block random performance, and lower power), the truth is that SSDs are still 3 to 30 times more expensive than hard drives. As a result, the most efficient cloud data centers use SSDs sparingly for only the hottest applications requiring the highest read/write performance.
HDDs continue to store the vast majority of data because of their cost/GB advantage. The smart focus is on selling both HDDs and SSDs in a tiered infrastructure. HDDs are most efficient for the vast majority of cloud applications in Tier 1 (10K/15K RPM SAS hard drives for mission critical data) and Tier 2 (7200 RPM SAS/SATA hard drives to store objects, social network content and Big Data), while SSDs are reserved for Tier 0 high performance transactional applications.
Step 3: Start with the Most Reliable Drives for Immediate Savings in CapEx and OpEx
Obviously the more reliable the drive, the less time and cost your customer spends maintaining it and the fewer returns you receive. Drives rated at the industry’s leading benchmark of 2 million hours Mean Time Between Failure (MTBF) will experience 40 percent fewer failures during the five year life of the drive over those rated at 1.2 million hours.
As an example of this MTBF impact, let’s consider a large deployment of 400K drives whose number increase 10 percent annually to reach a total deployment of 2.4 million drives in five years. During that five-year period, you would expect to see 20,000 more drives fail if there were rated at the lower 1.2 million hour MTBF than if they had the higher 2.0 million MTBF rating. If those failing drives were in a RAID configuration or a replicated environment there would be additional costs associated with performance degradation, while the RAID is being rebuilt, or by moving the data elsewhere in order to maintain system integrity.
Those costs are in addition to the costs of physically identifying where the drives failed and initiating replacements. At an estimated $100 cost per drive per incident, you’d be looking at spending an additional $2M by year five. If you’d selected drives with a three-year warranty, you’d have to pay for new replacement drives as well.
Step 4: Get the Most Out of What You Already Have: Move to 4-TB Drives
The TCO savings of moving from 1, 2 or 3TB drives to high capacity 4TB drives is substantial. Here’s why. Let’s say you want to build a video server with a 4U server case. Two SAS host bus adapters can connect 16 drives each for a total of 32 hard drives and the server can support a pair of SSDs for performance and a 10GB/s network card. Using 4TB drives in the 32 slots results in 128TB storage per server, while using 3TB drives results in only 96TB storage per server.
Let’s say your video library requires 384TB in storage. You would only need three servers using the 4TB drives vs. four servers using the 3TB drives. One less server translates into a 25 percent savings in power, and with 25 percent fewer rack components you could expect additional savings of 25 percent in cooling, 25 percent less rack space and 25 percent less weight.
Additionally, new up-and-coming helium-filled platforms are capable of supporting seven platters per standard 3.5-inch HDD, two more platters or disks than the current air-filled five-platter drives. As a result of the helium filling, the seven disk drives will deliver massive capacity while consuming 23 percent less power per drive, delivering 45 percent better power density and running four degrees Celsius cooler — reducing power and cooling requirements in the system rack and the data center for greater TCO.
Step 5: For Lowest TCO Learn from the Cloud Masters: A PUE of 1
Even a savings of a single watt per drive in basic power consumption across a large deployment over a period of five years adds up. If you deployed 12 hard drives per server and 25 thousand servers, you’d end up using 13 million kilowatt hours of electricity over five years. At an average commercial rate of 10.25 cents per kilowatt hour that single watt would translate into a cost savings of $1.3 million. By moving to 4TB drives and eliminating servers, battery backups and networking gear, the savings are exponentially higher.
Drive power consumption is one thing, but data center efficiency is even more important. Power Usage Efficiency or PUE refers to the ratio of the total amount of power used by a datacenter to the power delivered or consumed by its equipment. PUE helps you measure such factors such cooling, power distribution and lighting. The ideal PUE ratio is 1.
Best-in-class data centers such as Google and eBay have reported ratios as low as 1.14 and 1.35, respectively, but a typical data center has a ratio of 2.5, according to the Uptime Institute. So, a typical data center uses two-and-a-half times more electricity than the amount needed to run the actual equipment. A drive’s read/write and idle power usage figures indicate the power a drive requires to operate. Two terabyte drives have power costs between $36 to $48, depending on whether the drive is idling or actively reading and writing. For a 1TB drive the cost ranges from $48 to $100.
Using fewer, higher capacity drives will consume less power and make the data center more efficient. In an average data center with a PUE ratio of 1.5, every 4TB drive deployed instead of a 1TB drive will save an average of $111 per drive during the next five years. For a deployment of 10,000 drives that adds up to $1.1 million.
The bottom line is that there are multiple ways for you to profit by selling the value of storage as it impacts the overall TCO of the cloud data center. The factors your customers should consider include: using the highest quality drives; using the right mix of HDDs and SSDs by considering type of application and function across a systems-level, tiered architecture; and moving up to highest capacity drives to reduce basic power consumption and increase the efficiency of how that power is used.