“NVM” in all forms and shapes (SMC – ex. NVM, NVMe SSDs, NVME-OF), SDS crossing the “Chasm”, further consolidation in the industry – 2017 storage trends
Storage media and devices
Storage-class Memory (SCM), NVM previously
This was previously known as NVM (the storage media, not to be confused with NVMe/NVMHCI the interface), but the industry seems to have now settled on the term SCM. Having a dissimilar term was the right thing to do.
SCM is a new class of storage media. It is faster than (NAND) flash, slower than RAM, but it is non-volatile. The first products didn’t get released in 2016, however, we are now promised they will hit the shelves in 2017. Intel’s Optane products are the most developed of the bunch, but even for the details are still too scarce. The rest of the technologies seem to be in flux, with no concrete public plans, but lots of chatter about how great it will be when they are eventually out.
In the beginning, prices of SCM will be higher than the ones of NAND flash, but in the long run, an SCM technology can replace flash entirely. As a result, NAND flash is going to become the new disk, while SCM will take the role of flash. This means a new crop of storage vendors focusing on the new storage tier as it happened with NAND flash.
NVMe SSDs
NVMe (aka NVMHCI) SSDs are already on their way to become a mainstream technology. It has completely won over SATA SSDs in phones, tablets, and laptops. In 2017, there is good chance NVMe will become the predominant datacenter SSD technology, displacing SATA and SAS SSDs. It’s about bloody time if you ask us. We were excited to see the NVMe roadmaps of several SSD vendors. Bigger, faster, stronger drives ahead, at potentially the same or even lower price.
Note that all NVMe SSDs on the market today are using NAND flash media, while in 2017 storage trends we expect to see the first non-flash NVMe drives. More IOPS, more MB/s and more importantly lower latency.
Storage systems
SDS & Distributed Storage crossing the “Chasm”
SDS is finally getting a solid foothold on the market. It generally takes 5 to 8 years for a new product/solution group to get traction on the market. SDS has been buzzed since 2013, and we saw an inflection point in 2016 – SDS was the first alternative reviewed in new projects, before traditional SAN, and alongside stand-alone all-flash arrays.
We expect SDS and distributed storage to Cross the Chasm, following the model Geoffrey A. Moore set out in his book “Crossing the Chasm.” Thus far SDS was mostly used by the Innovators and Early Adopters as per the Technology adoption life cycle. As a result, the technology proved its benefits and become more mature. In 2017 the Early Majority will adopt SDS and we expect the technology’s market share will rise to more than 20%. In turn, this will drive even more adoption in 2018-2020.
Hyper-Converged Infrastructure/Systems, HCI
HCI is getting to the point where it is becoming a standard for use cases which require simplicity above all. It was widely pushed in use cases it didn’t belong, such as building infrastructure at scale. Now it is becoming more or less mainstream technology in our view and a prime choice for companies which:
- Do not have large enough scale (think ROBO for example – many locations with small IT needs) or
- Do not have the technical expertise, time or willingness to manage separate infrastructure layers by themselves
The natural tradeoff customers have to realize with HCI is that they trade simplicity for higher cost, lower flexibility, and higher vendor lock-in. For other use cases (larger scale of deployments, or where IT is the main business of a given company) the best solution is still to manage different layers of infrastructure, although with software being the main tool.
Because of the above, we expect consolidation in the HCI market in 2017 as it becomes more mature.
NVMeoF
NVMe over Fabrics (NVMeoF, NVMe-OF) is a new storage network protocol (think iSCSI, but way faster). It requires RDMA support in the underlying network. Hence a fabric is a network with RDMA support.
On the host side, the aim is to have standard initiator software, shipped with the operating system.
On the target side, there are different approaches. One approach is to pass NVMe commands straight to an NVMe device. This would be similar to an iSCSI or FC target passing SCSI commands directly to the underlying SAS device. It is done for speed.
The other approach is to run a full-featured storage system under the NVMe-oF target, hence gaining higher data management features, and potentially increasing latency to a point where the advantages of NVMeoF are obliterated.
The first NVMeoF products are in various demos and PoCs (Proof of Concept). How fast this market will develop is questionable, given customers are still absorbing the benefits of “good-enough” performance of current generation flash-based storage systems. So we expect to see it in some latency-critical use cases this year with wider adoption the years after.
Full-featured storage systems vs. speed demons
With the wider adoption of NVMe (NVMHCI) and NVMe-OF, storage systems will become increasingly separated into two categories, with the gap between them widening:
- Super-fast systems with virtually no features. Think of them as flash JBODs or “JBOFs” for short.
- Feature-full storage systems with all the data protection and management capabilities expected of good SANs (end-to-end data integrity, volume management, fine-grained snapshots, thin provisioning, scale-out, API control, DR)
Very few vendors will manage to ship a stable storage system which is both feature-full and fast in this new environment.
Consolidation in the storage industry
We expect further consolidation in the storage industry. There were too many “me too” vendors, heavily funded by venture capital (VC) in 2013-2016, but with weak products. And while you can buy a lot of marketing/coverage and mind-share with VC money, you cannot make a customer buy or keep a product that does not work well.
So, on one hand, we expect there will be less new vendors started in 2017 and on the other hand we expect further consolidation of the incumbents as we witnessed with Dell/EMC. We already see that better products make a name for themselves, regardless of the size or initial brand awareness of the vendor. The reason is customers vote for quality products which advance their technical capabilities and deliver measurable business value.
Storage is going to become even more exciting and diverse in 2017. So keep your mind open and look for the facts and underlying substance and not the marketing “buzz”.
To be continued with “IT Infrastructure Trends and Predictions for 2017“. Stay tuned!
Edit: You can now review the “IT Infrastructure Trends and Predictions for 2017”
If you have any questions feel free to contact us at info@storpool.slm.dev
Share this Post