First developed in 1984 by Toshiba flash memory is an electronic non-volatile computer storage medium that can be electrically erased and reprogrammed. The NAND chip is commonly found in digital devices such as digital media players and cameras, smartphones and more recently on personal computing devices such as tablets and laptops. It is used in main memory, memory cards, USB flash drives, solid-state drives, and similar products, for general storage and transfer of data.
When the considerations are for speed, noise, power consumption, Flash memory, in the form of solid-state drive (SSD) has become a popular replacement component for hard disks in laptops. These features have not been ignored within the enterprise where vendors have started to incorporate NAND technology as replacement or complementary storage to existing hard drive-based storage arrays and enterprise servers with RAID and SAN architectures.
Initial objections to the use of flash-based SSDs as enterprise storage, including high cost per gigabyte of flash memory and finite number of P/E cycles, are rapidly in decline as warranties on flash-based SSDs are approaching those of current hard drives. It also helps that vendors are continually making innovations in software and storage design architectures to take advantage of the characteristics of NAND technologies.
Data&Storage Asean interview three different proponents of NAND technologies as used in enterprise storage to get their particular why and how their technology is gaining traction in Asia. To be clear, there are currently three approaches to NAND technology use in enterprise storage.
Violin Memory designed and built storage appliances using a backplane-based approach that directly incorporate NAND chips. The company argues that the choice of which SSD a vendor makes in designing their array is secondary to the design choice to use SSDs or flash chips directly. The vendor created a custom architecture called the Flash Fabric Architecture to use flash chips directly. One key aspect of flash design is performance, and SSD-based vendors typically suffer from hot spots in performance, whereas Violin’s Flash Fabric Architecture allows flash management/garbage collection, to be managed on an array-level, not a SSD-based level for a more efficient design. This avoids unpredictable spikes in latency found in SSD-based designs. Violin’s approach delivers consistent performance, even on heavily loaded systems.
Another aspect of Violin’s design is density. By using a backplane design the vendor is able to cram 70 raw TB on a 3RU space with power consumption of about 1500W. Violin claims SSD-based designs would need to pack a similar capacity solution into a 12RU and consume 3000W of power. One customer estimated they would save 71 percent in operating costs with Violin over traditional storage.
Violin Memory says their solutions are particularly designed for mixed and multiple workload environments with a variety of applications that deliver ‘transformative performance’ with lower total cost of ownership (TCO) than legacy or SSD-based storage.
Pure Storage took a different taking what is today a commercial success – solid state disks or SSDs – and building an external storage array or appliance around it, and complementing it with advanced software technology to optimize the new device for the data center. In its report “Market Share Analysis: SSDs and Solid-State Arrays, Worldwide, 2013”, Gartner noted that Pure Storage took a fresh and simple approach predicated on compelling data management software features.
According to Michael Cornwell, Pure Storage chief technology officer, Asia Pacific & Japan, “where the IBM flash systems use inline compression and the EMC flash products use inline deduplication, Pure Storage has broken the economic barriers of flash adoption in the data center by being the only AFA with both inline compression and deduplication shipping today. This allows users to store more data for less on flash.”
On 19 June 2014, Pure signed a cross-licensing agreement with IBM giving it access to more than 100 Big Blue IP around storage. Pure Storage CEO Scott Dietzen says this is his company’s strategy to protect itself against potential IP-based litigation from competitors.
Echoing the same sentiment, Cornwell adds “going forward, our hope is that this significant expansion of our patent portfolio will serve as a clear warning that Pure is serious about defending our ability to disrupt a storage marketplace that is deeply in need of innovation, and moreover, that Pure is in this for the long haul.”
On the hardware front, Pure continues to defend its strategy of using commercial-grade NAND chips, called cMLC, to build its products citing that due to the extreme cMLC volumes today, the technology actually has the lowest failure rate among the different types of NAND, including SLC, MLC and eMLC.
Pure Storage strongly believes that consumer-grade MLC flash is key for both achieving the price-point that will make flash disruptive in the enterprise, and for delivering the most reliable all-flash array. “We have chosen to leverage the most cost-effective consumer-grade MLC flash available, and to build reliability technologies into our core Purity Operating Environment software, enabling us to deliver these services more efficiently globally instead of relying on expensive eMLC SSDs,” said Cornwell.
He also argued that if an all-flash array leverages eMLC or SLC flash, it is a ‘canary in the coal mine’ that its respective vendor simply hasn’t done the software work to properly understand and address flash, instead choosing to rely on exotic enterprise-grade SSDs, a path that will ultimately prove to be un-economic and less reliable.
Which leads us to NetApp, which in this article, we chose as representative of traditional enterprise storage array vendors, the likes of EMC, IBM, and HP, all of who are now releasing what they claim are All Flash Arrays (AFA) – but in my humble opinion – are simply traditional storage arrays with their HDD guts taking out and replaced with SSDs which analysts and vendors label Solid State Arrays (SSA). Of course as these vendors rush out to announce their AFA offerings, they are also fine-tuning their software to take advantage of the inherent characteristics of flash technology.
According to Low Kai Chin, District Manager for NetApp Malaysia and Brunei, “We have seen a growing demand across Asia of late for all-flash arrays, especially from the larger enterprises that need storage and data management solutions that have the scalability for short-term efficiencies and long-term growth. However, we would like to stress that despite the hype and popularity, there is no ‘one-size-fits-all’ approach to flash. Enterprises need different features: performance, efficiency, reliability, flexibility and choice to maximize the value of flash.”
NetApp believes that at this stage of the flash revolution, it is really a question of which format can offer the right balance of price, reliability and performance. In some cases, transactional workloads such as database, OLTP, web services, and virtual desktop infrastructure (VDI) can run 500 percent faster than in traditional storage environments.
“At the same time, there are scenarios where all-flash solutions with a single active controller can potentially become the performance bottleneck. We believe that the landscape will continue to evolve and mature, making it difficult to determine the dominant format especially when customer needs and demands continue to evolve as well,” said Low.
Incidentally, NetApp ranked fifth in the Gartner “Market Share Analysis: SSDs and Solid-State Arrays, Worldwide 2013” report behind IBM, Pure Storage, Violin Memory and EMC. NetApp dropped from its third place spot in 2012. Gartner estimated total vendor revenue for SSAs in 2013 at US$667 million up 182 percent from 2012 levels.
Gartner predicts that developed markets will continue to lead in the acquisition of SSAs with China being the notable exception, outperforming others by virtue of the presence of web-scale, e-commerce and other major data centers. The traditional storage vendors will continue to lead the market in Asia because of their established physical presence, mature distribution partnerships and mind share.
Said Cornwell, “Asia has been a big contributor to our record global growth; we have been surprised by the rapid adoption of flash storage. A lot of this adoption is driven by the fact that Asian technology buyers are really motivated by return on investment. We see demand from all over the region. Singapore is great market for us as many multinational and financial services companies have relocated their IT leadership and technology decision makers to Singapore. Customers we are engaged with in Singapore often are not only making the storage decision for Asia but for the rest of the world. But we also see demand from non-traditional markets: Vietnam was one of the first markets where we started to see demand from customers while New Zealand customers are swapping out disk for flash at an impressive rate.”
Emerging markets definitely have the ability to learn from the technology decisions that have been made by their more developed neighbors. “Late-comers” to flash storage technologies will be able to enjoy the advantage of more use cases of early entrants using innovative and interesting ways to achieve greater performance and agility in order to drive their business forward.
Cornwell claims that educational institutions are benefiting from flash arrays opting to use the technology by deploying virtual desktops as opposed to purchasing and providing computers for students. He also says that Infrastructure-as-a-Service providers stand to benefit from flash.
“Asia possesses the communication infrastructure to enable enterprises to consolidate their data center infrastructure without significant costs, with managed service providers in the region. These managed service providers are able to deliver better SLAs at a lower cost to the customer with the emergence of all-flash managed services, which we help power. Flash storage with the benefits of low latency and data reduction at the price of legacy storage is a true differentiator. Our customers are seeing a true competitive advantage,” Cornwell concluded.
Low says his company’s strategy is simple. “We want to provide the best value for our customers. They seek performance, but not the cost of being expensive or less reliable. Thus, we believe that in today’s environment, these requirements are best met by hybrid systems. We have the strongest and broadest flash portfolio in the industry to provide flexibility and choice to maximize the value of flash across the entire compute, network, and storage stack. We know flash. After all, we bring to the table a deep knowledge of enterprise workload needs with more than 20 years of innovation in storage and data management,” said Low.
Violin Memory’s response to our question around marketing strategy is to let the customers speak for themselves. In this case, the vendor listed Sinopec, Korea Exchange, GAMEVIL and iTalkBB as among its growing customers in Asia.
For many applications and enterprise data centers, flash is no longer the nice to have technology that takes too much effort to justify. Why? Because the debate has shifted from dollars per Gig to IOs per Gig! It’s now a cost of doing business and whether the current infrastructure is able to meet the performance requirements of business.
What is true though is that the discussion has become far more complex and all encompassing. The CIO and his team needs to look at each application that Lines of Business use, project the demand versus the capabilities of the data center, and then map it into a cost analysis table to determine the path the organization must make to stay competitive.