Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Enterprise SSD technology seems to get more complex over time. There are all the types of cells, 3D varieties and a host of form factors. Some may hanker after the good old days when an SSD meant a solid state drive you could use as a direct replacement for a disk drive. But the options available today are many, and certain types are best for certain applications.
Here, then, are some enterprise SSD tips regarding form factors and general flash implementation.
No One Champion
It’s doubtful that there is “one form factor to rule them all.” Some may be best for X but lousy for Y. But Greg Schulz, an analyst at StorageIO Group notes that many vendors attempt to champion a particular flash SSD form factor and interface, claiming it’s the best and only fit for the enterprise.
“The reality is that there are different options from PCIe to M.2 to U.2 and traditional drives for different needs,” he said. “It really comes down to having multiple tools in your SSD toolbox to use for different needs. Granted if all you have is type of SSD, you can guess what the answer to any situation will be.”
Some advocate flash everywhere for everything. And the reasons are simple: the technology is mainstream, the performance benefits are well-chronicled, and the acquisition cost concerns are largely in the past. As a result, Dell EMC has made enterprise SSD an integral part of its many storage platforms and applies it to every class of workload.
“Gone are the days where flash was only applied to VDI or the most performance-critical applications,” said Dan Cobb, vice president of flash storage strategy and fellow, Dell EMC. “Flash now applies across small, medium and large data centers and for block, file and object data.”
The IEEE Computer Society made some technology predictions for the coming year. Its main comment on storage was to watch for the rise of fabric-attached Nonvolatile Memory (NVM). NVM may have enjoyed mixed success in productization to date, but the number of companies working in this arena makes it a likely candidate for rapid adoption.
According to IEEE, fast, nonvolatile storage bridges the gap between RAM and SSDs, with a performance-cost ratio lying somewhere in between. It can be accessed by the operating system (OS) like any other permanent storage device or deployed in DIMM slots, accessed by the OS as memory.
Some vendors use different form factors for different use cases. Tintri, for example, favors flash everywhere — serving out LUNs over iSCSI, file systems over NFS or SMB, or object over S3. But it uses different flash form factors for scale-up and scale-out.
“For scale-up architectures, we favor SAS SSDs, which is the most common approach today,” said Craig Schultz, director of product management, Tintri. “but we expect the industry to move to dual-ported NVMe SSDs over the next two or three years.”
For scale-out architectures, Tintri takes a different view. In that case, it recommends SATA SSDs to keep costs low. Its Micron solution, for example, is based on SATA SSDs. Over time, though, that is likely to change as new form factors gain adoption and maturity.
“Things are currently shifting to scale-out all-NVMe clusters for scale-out, with low-cost single-ported NVMe devices in each node.,” said Schultz.
The right enterprise SSD platform isn’t just about the hardware; software is rapidly becoming a real source of differentiation. That includes analytics that help you predict your exact need for storage capacity, performance and working set. So don’t forget about such functionality when you are adding flash. Some vendors are further ahead than others when it comes to flash analytics.
“Advanced analytics allow ‘what-if’ scenario modeling, where you can determine the impact of changes to your environment before implementation — that’s how you avoid surprises and unnecessary costs,” said Schultz.
Michael Letschin, field CTO, Nexenta, said to consider choosing software-defined storage (SDS) to extend the benefits of all-flash to a wider set of workloads. He said that with SDS and the right architecture, you can achieve tremendous total cost of ownership (TCO) savings, for tier 1 and even tier 2 applications.
“We’re seeing customers going all-flash at multi-PB scale, taking flash to all active workloads. SDS on all-disk should only be used for backup and archive use cases,” said Letschin.
Letschin added another tip. When it comes time to add capacity and/or performance, that’s where expandable units that work at the virtual machine (VM) level prove their value. Traditional block storage requires that you find room in the rack, add a shelf (or two), connect cables, carve additional capacity into LUNs and assign them to virtual machines.
“When storage operates at the virtual machine level, you can expand capacity in minutes and make it immediately available to the hypervisor,” said Schultz.
Up and Out
Some argue that scale-out flash is best. Others make a good case for scale-up. So who is right? Neither has the right answer believes Shai Maskit, director of technical marketing, Kaminario.
“The form factor of flash storage for the enterprise is an all-flash array that can scale both out and up,” said Maskit. “The advantages of shared storage outweigh those of direct attached: better resiliency and better utilization of the storage, making it the best choice for the enterprise. He added, “As the enterprise environment grows, you need to be prepared to add more flash storage in the most cost-efficient way. Choose an array that can scale to meet your demands of capacity and performance, not only for today, but for tomorrow.”
Photo courtesy of Shutterstock.