point
Menu
Magazines
Browse by year:
August - 2016 - issue > Cover Story
Nimble Storage Delivering Data Velocity
si Team
Monday, August 1, 2016
February 23, 2016, the One Kearny Club at San Francisco, CA was bustling with the presence of storage industry fraternity. The star of the show was a burgeoning, eight year old company, Nimble Storage (NYSE:NMBL). But, the prominent faces and international media presence-along with around 1500 pairs of eyes glued to the live streaming-made it clear that the event was something worth witnessing. Occupying the stage was Suresh Vasudevan, the technology aficionado born and bred in India, nurtured by the U.S. who took over the reins as CEO of the company within three years of its inception. Within two years of taking charge, he led the company to IPO. Trained and sharpened at one of the top technology institutions and IIMs of his homeland, this CEO proved his mettle when he transformed a six million dollar business into $60 million venture within 365 days during one of his previous stints. Vasudevan and his team at Nimble Storage foresaw the future of storage in flash drives and leveraged it when the legacy powerhouses did not-a time when established players like EMC2 retrofitted flash as bolt-ons in their disk drive-centric storage products. With much deliberation Nimble stepped into the market with a software architecture designed to leverage flash-a disruption as well as complimentary force to hard drives-which allowed them to tap in best of both the worlds. With this Nimble managed to add the extra horses through flash derived performance required to attain momentum in the wake of big data dump slowing down enterprise applications. In the meantime, flash's other-half-hard drives-maintained the cost-efficient medium to gobble up the big data thrown at it.

However, two decades of experience in the Silicon Valley taught Vasudevan the fate of most start-ups; they come under the radar of giants who are ready to acquire budding competitions. Banking on flash drives, he knew technology was not the culprit. Flanked by an experienced team, Vasudevan understood that the problem lay with the lack of scalable after-sales support that mandated customers to monitor the systems themselves. Any issue resulted in a cascade of troubleshooting between the customer and vendor, starting form Level 1 engineers; resolution was resource intensive and typically required days. Based on this conclusion, Nimble team stumbled upon storage industry's next innovation: eliminate the need for customers to monitor their systems and prompt diagnostics; leverage cloud and predictive analytics to monitor a customer's environment and deliver them prescriptive recommendations about what might go wrong. The gathering at One Kearny Club was present to witness the result of Vasudevan's and his team's brainstorming-the industry's one and only predictive all flash array, Nimble Predictive Flash Platform. The Nimble Storage Predictive Flash Platform is designed to improve the performance of a client's analytic applications and big data workloads and meet cost and capacity SLAs by leveraging predictive analytics and a unified flash fabric. "The Predictive Flash Platform combines flash performance with predictive analytics to close the app-data gap even when the barriers to data velocity are not storage related," informs Vasudevan.

Bridging the App Data Gap

"Simpler is faster", and Vasudevan believed simplification of the architecture could be a key to the market that has long been dominated by big players whose architectures were complex. He knew customers care for experience of their applications, and complexity was a hurdle. "It's clear that complex infrastructure creates an app-data gap that slows down data delivery which impacts critical business processes," says Vasudevan. Backed by a simple architecture, the sensors built in Nimble OS software and physical sensors embedded in storage and the surrounding infrastructure scattered across the globe made a critical revelation. Through the 30 million pieces of health data their sensors transmit per system per day, it was clear that directly storage related issues constitute a big chunk of App-Data Gap-around 45 percent. But, the rest 55 percent arises from the remaining infrastructure stack, such as networking, VMs, and non-adherence to best practices. Flash drives that are about 100X faster than hard drives solved most of the percent challenges in narrowing the gap. Vasudevan and his team knew tackling the rest 55 percent can be a game changer and the remedy came in the form of InfoSight, the predictive analytics engine built into every Nimble array ever shipped.

Combining their huge database of sensor reported health information and cloud-based predictive analytics, Nimble unraveled the hidden window to monitor and proactively deliver recommendations to the customer, before the issues disrupt the mainstream applications. Rod Bagg, VP, Analytics and Customer Support, Nimble Storage, says, "Here's our idea of big data: InfoSight collects more anonymized sensor data points every four hours than there are stars in our galaxy. At over 200 trillion sensor data points collected a year, we have amassed what is probably the biggest repository of storage-related sensor data ever collected." He goes on to narrate a customer's story. Facing a problem within the application infrastructure, which was not storage related, the customer contacted Nimble after six weeks of unresolved troubleshooting with the concerned vendor. Nimble solved it the same day. Thanks to automated data that eliminated the need for Level 1 and 2 engineers, health data of the entire stack and InfoSight.

Predictive Analytics Delivered

The Predictive Flash Platform promises to be the solution the industry seeks, reeling under a monstrous appetite for performance and capacity, and the need for analytics fuelled by big data. Paraphrasing Vasudevan, the platform is built upon three components: InfoSight, Unified Flash Fabric (UFF) and Timeless Storage. The UFF is Nimble's OS architecture that brings together their latest all flash arrays and long-standing line of adaptive flash arrays under a consolidated platform to accelerate enterprise applications and lower the total cost of ownership (TCO). A unique market innovation under UFF is Scale-to-Fit. With this scaling technology, customers can start with a single all flash array, with the option of adding additional all flash shelves to multiply the capacity when needed. It also allows performance up gradation by replacement of controllers with more powerful ones offering the power of extra processors and DRAM. For customers exceeding the requirements of a single system there is Scale-Out Clustering technology that delivers the combined power of four systems, seamlessly. The results are simply astounding. While a single array can deliver 2 PB of performance or 300,000 IOPS, clustering offers 8.2 PB or 1.2 million IOPS.

Another unique feature under UFF's belt is the Mix-and-Match capability, which allows customers to leverage adaptive and all flash arrays within one Scale-Out Cluster. This allows deploying and switching applications between adaptive and all flash arrays based on the customer needs at the push of a button. Owing to the consolidated operating architecture developed by Vasudevan and his team, Scale-to-Fit and Mix-and-Match are highly performance intensive and non-disruptive to application operations.

Another feint where Vasudevan has excelled is dramatically lowering the TCO. Data reduction making the lineup 10 to 30 times memory efficient compared to other products; metadata footprint optimized system freeing up more DRAM making the controllers less expensive; and cost-effective dense 3D NAND and unique endurance management translate to a 33 to 66 percent less TCO compared to other leading products available. What is remarkable is the non-stop availability awarded, with an actual measurement of 99.9997 percent, which accounts to less than 95 seconds of unplanned downtime per year.

All this with less than one millisecond of latency across entire enterprise and applications acceleration by five to ten times. In Nimble's long list of innovations lies Triple+ Parity Raid, which provides operational tolerability of up to three SSD failures and the benefit of intradrive parity offering the highest degree of resiliency in the industry.

Summing up, Nimble's all flash arrays leapfrog ahead of every other product in the market in terms of scalability, lowering TCO, and availability and resiliency. The compounded result of these benefits is a tool that enterprises can leverage to extract actionable insight from their big data powered by innovations that are shielded by state-of-the art data protection.

Performance viable technology

Vasudevan believes, that although flash is gradually displacing hard drive, the latter still has a considerable mileage left and cites the main reason to be the relative cost-endurance characteristic of present flash technology. While there are currently flash drives available suited for mainstream enterprise applications from an endurance standpoint, their high cost inhibits a large-scale deployment. Moving down the cost curve, there are less expensive flash drives in the market, but their endurance is appropriate for consumer devices, not enterprise deployment. The advent of public clouds has been a worry for the industry and Vasudevan admits that in the longer run it will affect every player in the space. "Currently the growing market share of flash optimized storage technology, lies between 60 to 70 percent and a reducing share of the hard drive counterparts to 30 to 40 percent gives us the advantage," said Vasudevan. The pace of its customer accumulation can gauge Nimble's rapid rise, 7500 customers spanning in over 50 countries. Judging the well planned timing to enter the market and introduce new products, and the cards that Vasudevan has played so far, it can be assumed that Nimble's expansion plans, and the aspiration to secure the top spot looks within the grasp.
Twitter
Share on LinkedIn
facebook