Evolution of Enterprise Flash
If we look at a few years ago, the flash was considered only for individual use cases and used for sub-millisecond response time requirements. Further, the flash prices started dropping steadily, and many vendors are selling flash drives and flash arrays for prime workloads. The cost for the flash drive is almost equal cost to spinning media’s value and then the reliability factor of the drive itself increased the adoption. The efficient garbage collection process, meaning the writes are handled in the way of load balanced across the storage medium to increase the lifetime of the drives. Flash becomes the mainstream, and it is part of the conversation with the customer whether it is enterprise or off-the self hardware. On the other side, the wave of the public cloud adoption continues to show the impact to the every enterprise storage vendor’s business. IDC projected when the customer spends $1 on the public cloud storage services, the enterprise storage market share declines up to ~$3. Storage solutions overall have to come up with the unique value proposition to show the differences and provide the options for customers to choose the right platform for their workload. Let us discuss Flash memory-based architectures, NVMe and PCIe SSDs, and controllers are innovating and delivering the high performing workloads” and how it is helping enterprise customers.
NVMe Based Flash Architecture
NVMe is the new standard protocol for talking to flash memory. It appears that every enterprise vendor is incorporating NVMe into its products, although NVMe platform yet to prove in the industry. NVM express or nonvolatile express controllers is a logical interface accessing non-volatile storage media attached via PCI Express bus. In detail, NVM express allows the host hardware and software to increase the possible parallelism level in modern flash devices. As a result, the NVM Express reduces the I/O overhead and brings various performance improvements in logical-device interfaces, including multiple, long command queues, and reduced latency.
The next wave of NVMe is NVMe over Fabrics” Enables Storage Networking of NVMe. Sharing NVMe-based storage with multiple servers. This provides Better utilization: capacity, rack space, and power, Better Scalability, Management, Fault isolation. According to IDC, 50% of enterprise servers and storage appliances will support NVMe by 2020, ~40% of all-flash arrays will be NVMe-based by 2020, Shipments of NVMe SSDs will grow to 25+ million by 2020
Impact of creating New Storage Tier: With the introduction of NVMe
NVM Express reintroduces the Tier 0 with the characteristics of 1,000,000 IOPS and latency up to 100 Micro seconds. The new Tier1 changed to 100,000 IOPS with sub-millisecond latency, and other tiers continued to be there for few more years and expected to phase out.
Tier0- NVMe Flash - Drives 1M IOPS; 100 Micro seconds
Tier1- SAS/SATA SSD drives 100K IOPS; sub-millisecond latency
Tier2- When NL-SAS drives prices are equal to the Quad Level Cell Flash and this option could replace this tier completeley
The primary workloads for new flash are running databases, machine learning devices, Artificial Intelligence devices, and anything data that provides the important process on fast reads, writes and lower cost. Software Defined Storage (SDS) is seen as the future in the storage ecosystem, and NVMe expected to have the pure play in providing the extreme performance in features like automatic data classification and prioritization based on usage or access patterns, automatic defragmentation, optimization, compression, replication. Flash will dominate the storage solutions on all emerging applications. NVMe and NVMe-over-Fabrics would help to enable the flash adoption on prim and continue to evolve in Public cloud services and reducing further Hybrid
Next one is coming:
How machine learning and AI is going to help driving the use cases for the storage offerings.