The new 3D NAND SSDs are designed for cloud data centre scenarios, but those conditions apply to more than cloud service providers, according to Intel.
Intel has rolled out its latest "cloud inspired" SSDs for data centres as the tech giant continues to push forward the 3D NAND option and the NVMe standard.
Jonmichael Hands, a product marketing manager at Intel, said the company's two new 3D NAND SSDs were specifically designed for cloud data centre scenarios, but that those conditions apply to more than cloud service providers. In a telephone interview with EE Times, he said that although many organisations are spending more money on the public cloud, “at the same time, enterprise IT is moving toward the private cloud.” Either way, they are seeking performance and efficiencies.
Intel's 3D NAND SSDs are aimed at converged infrastructure deployments where capacity is important, but also flexibility so that enterprises can easily re-task systems for new workloads, as well as improved both CAPEX and OPEX, said Hands.
The DC P4500 line is optimised for reads and is aimed at helping data centres get more value out of servers and store more data, said Hands, while the DC P4600 series accelerates caching and enables more workloads per server. Both use Intel's TLC 3D NAND combined with a new, Intel-developed controller, new firmware and PCIe/NVMe. Both series will initially come in a half-height half-length add-in card and U.2 2.5-inch form factors in 1, 2 and 4TB capacities.
Intel's new firmware includes support the NVMe-Management Interface (NVMe-MI) specification, said Hands, and offers manageability features to reduce server downtime through improved update processes and expanded monitoring capabilities, with management coverage is now expanded across a wider range of drive states. The new controller, meanwhile, supports scaling as high as 16TB, he said. “That's really only limited by the amount of flash you can fit in that form factor," he added.
Hands said Intel is confident in the forward momentum of NVMe, noting that both Microsoft and Facebook have seats on the NVM Express board, and cited its scalability, open standards and flexible form factors as attributes contributing to adoption. “The industry has chosen NVMe," Hands said. "It is the next protocol for SSDs."
Hands added that all new SSD deployments in the hyperscale segment are NVMe, although the overall enterprise market is slower. “The inflection point is happening," he said.
Figure 1: Intel's DC P4500 line is optimised for reads and is aimed at helping data centres get more value out of servers and store more data, said Hands, while the DC P4600 series accelerates caching and enables more workloads per server.
Hands said the SATA to NVMe transition is inevitable, as the older interface is becoming a bottleneck, while NVMe higher performance enables systems to get better value from the flash. “We can take full advantage of the 3D NAND," he said. “You're exposing it on a interface that's been designed for SSDs."
Intel has been expanding Fab 68 in Dalian, China, to increase its 3D NAND supply to better meet the storage needs of customers.
Intel initially developed its 3D NAND technology in collaboration with Micron, opting to leverage floating gate technology that has been easier to manage than the problematic charge trap approach being used by everyone else, said Jim Handy, principal analyst at Objective Analysis. “Because Intel and Micron chose not to use charge trap in their flash, they've been able to ramp it faster than everyone else," Handy said.
And although the two companies have joint product facilities, their arrangement gives them liberty to build their own, said Handy. “Intel is ramping up production ability, enabled by the success of its 3D production with Micron," he said.
He said pretty much all hyperscale data centres are using flash in form or another. “They see it as a path to getting the most performance for their dollar," Handy said. In most cases it's a combination of flash, DDR and spinning disk, with the latter is still be used for cold storage, he said, even in hyperscale systems.
In the long term, it will be planar NAND that becomes the odd man out as 3D NAND costs go down. “There's a real limit for how many applications that will be suitable for planar NAND," said Handy. While it's possible it could displace NOR flash in some applications, there will barriers as its necessary to license sophisticated software.
And although it doesn't make sense to put 3D NAND in small USB thumb drive, it's making its way into endpoint devices. “The limiting factor is the smallest 3D NAND chips you can buy are 128Gb or 16GB," Handy said. "Anything that's got less than 16GB in it is not going to use 3D NAND."
First published by EE Times U.S.