In my last post, I argued the correlation between the value of data and its age has flipped in the last decade. As a result, IT strategists have had to question – and often reject – the cozy assumptions that progressively removed aging data from fast, front-line apps into slower, lower-cost storage options.
To be fair, devising an effective storage strategy has never been a simple assignment but the complexity of the task is so much harder today. That’s because predicting the future is never simple and, when it comes to estimating the future value of data, it’s just guesswork. Simultaneously, the range of storage technology options available continues to evolve and become more complex. Devising data-driven transformation – which Fujitsu sees as the core of digital business investment value – requires both a wide choice of storage technologies and a firm grip on what to use when in order to provide flexibility, resilience, and value.
Perfectly adapted for its environment
Last time I described the jungle of options available and how Fujitsu offers a combination of the best data storage technologies, including traditional RAID systems and more modern approaches like software-defined storage (SDS) and hyper-converged infrastructure (HCI), supported by data protection solutions like backup and archive.
Each of these approaches has distinct attributes, adapted to different circumstances. Extending this adaptation metaphor, we can think of each option as a unique species of tree. In this blog, I’m focusing specifically on traditional RAID (which stands for redundant arrays of inexpensive disks) storage. At Fujitsu, we think of these storage systems like the giant redwood trees of the storage jungle.
These magnificent trees are some of the oldest and tallest in the world. They can live for over 2,000 years and can tower over 300 feet. Traditional RAID storage is like the redwoods, continuously evolving to support both legacy enterprise applications and modern cloud-native applications, while effortlessly growing with massive scalability up to 100s of PBs to match the demands of scale-up and scale-out storage for rapid business growth.
The RAID terminology goes back a long way – maybe not from the perspective of redwood, but certainly for the IT industry. It was first used by David Patterson, Garth A. Gibson, and Randy Katz at the University of California, Berkeley, back in 1987, who argued that inexpensive drives developed for personal computers could outperform the top-performing mainframe disk drives of the time. Reliability would not be as high but, they argued, by factoring in redundancy, the overall resilience of an array could far exceed that of any large single drive.
The concept took off and today RAID storage is a pillar of data center infrastructure that has evolved to support both legacy and modern cloud-native applications. Naturally, the “inexpensive disks” of the original formulation are no longer recognizable, usually replaced now by solid-state flash memory SSDs. Expensive at first, SSDs were initially added to HDD arrays as a performance-enhancing layer to create hybrid arrays. This approach is still cost-effective today in many scenarios to balance capacity, price, and performance. Technology evolution and cost reductions then brought enterprise-grade all-flash arrays (AFAs) that are now normal for high-performance primary storage for optimal price and performance. The rapid maturation of the latest NVMe (non-volatile memory express) all-flash storage for extreme performance with minimal latency for the most demanding applications means that everything is changing again.
Hybrid, all-flash, or NVMe?
As usual, choice creates its own complexity and deciding which of these three types of an array to deploy requires thought and expertise.
Fujitsu’s approach is described in one of the ‘flashlights’ I mentioned in my last post, ‘Choice in a Shifting Landscape’, which helps cut through that jungle of complexity by showing business teams how to focus on and rate five key attributes for each workload. By scoring each of the five attributes on a one to four scales, this methodology connects the technical and practical needs of the business and identifies the right balance between hybrid, all-flash, and NVMe.
As a very rough guide to performance, where an all-flash SSD might be at least five times faster than a disk drive, an NVMe SSD could be five times faster than all-flash. But NVMe is about much more than speed. It has the potential to change almost everything, especially once the data path is end-to-end NVMe. Its support for massive parallelism and multiple threads provides better resource utilization and potentially moves the bottleneck back from storage to the server layer. While NVMe ultra-high-speed connection interfaces for memory-speed devices primarily exist “within the box,” NVMe over fabric (NVMe-oF) extends the very same capabilities across the entire data center and allows larger data sets and long-distance clustering while retaining key features such as QoS management and defined data locations for compliance.
Working out exactly when to deploy NVMe is not straightforward and our flashlight “Storage in the Era of NVMe” gives you the tools you need to make the right choices.
Optimize storage efficiency
The analogy between RAID storage and the giant redwood has other fruitful aspects. Like redwoods, RAID storage can help cut CO2 in the atmosphere. RAID storage can result in five to ten times less storage needed with compression and deduplication, and all-flash storage brings greater storage density, optimized capacity, and a lack of moving parts that means lower power consumption.
And in the same way that redwoods provide a habitat for a host of other plants and animals, RAID storage is ideally adapted for different interfaces, such as SATA, SAS, and NVMe.
You can find out more about Fujitsu’s approach to RAID storage and explore the data-driven storage jungle here.