It’s time to gaze into the crystal ball and think what might be on the storage agenda for 2019. What will the biggest storage trends for 2019 be? Let’s start with some easy ones.
NVMe over Fabrics will be big
Of course it will. All of the major storage vendors are already working towards supporting NVMe-oF and we have a slew of new vendors developing products around the new faster protocol. However, the question is, where will the initial growth and usage be seen? Based on our podcast discussions in 2018, we can expect that 2019 will most likely see adoption of FC-NVMe by enterprises looking to gain performance improvements out of their existing infrastructure. After all, not every application will need a high degree of performance improvement. Expect to see NVMe-oF based on Ethernet and InfiniBand being initially used by niche applications that have specific performance requirements. Here are some blog posts and podcast episodes covering NVMe.
- Pure Storage – Seeding the NVMe Market
- NVMe over Fabrics – Caveat Emptor?
- The Race towards End-to-End NVMe in the Data Centre
- Has NVMe Killed off NVDIMM?
- #76 – Fibre Channel and NVMe with Mark Jones
- #61 – Introduction to NVM Express with Amber Huffman
- #59 – Ethernet vs Fibre Channel
QLC NAND will complement TLC
QLC will not be a replacement for TLC flash because the performance characteristics are different. Up to TLC technology, the standard read/write ratio in the data centre was manageable with improvements in endurance management. QLC however, will be different and be more suited to read-focused workloads. This means we will see QLC and TLC alongside each other in platforms and solutions. The challenge for vendors and solution providers alike will be to ensure data is placed on the most appropriate tier of storage. This means we’re getting back to the says of storage tiering again. Some things never change. Expect to see a wider range of QLC devices in 2019, with some pretty big capacities – 32TB upwards. The key question for 2019/2020 will be what impact this has on the HDD market. QLC will be the most affordable flash storage, displacing a lot of 10K HDDs and opening up read-only processes like analytics to a wider user base. Here are some blog posts and podcast episodes covering QLC NAND flash.
- QLC NAND – how real is it and what can we expect from the technology?
- Micron users in the era of QLC SSDs
- SanDisk 400GB MicroSD Card – QLC in Action?
- Making The Case for SATA SSDs
- #79 – Learning More About QLC Storage with Steve Hanna
- #54 – Are we at All-flash and HDD Array Price Parity?
- #34 – NAND Shortage? What NAND Shortage! with Jim Handy
The performance battleground will move to the server
Although NVMe-oF will dramatically improve I/O performance, focus will move to the server where network delays can be mitigated or eliminated. If Persistent Memory is to be a success, then most I/O will have to occur in the server, or performance gains will be lost. The question to be answered in how these solutions work. Expect to see more success from the tiered model (think NetApp MAX Data), the caching model and scale-out distributed storage (e.g. WekaIO, Excelero). Here are some related podcast episodes and blog posts.
- Accelerating Workloads with NetApp Plexistor
- What are Storage Class Memory (SCM) and Persistent Memory (PM)?
- #52 – An Introduction to WekaIO Matrix with Liran Zvibel
- #48 – Introduction to Datrium DVX
- #36 – The Persistence of Memory with Rob Peglar
Now let’s think about problems from a customer perspective. What needs to be solved?
Data mobility will be a big challenge in hybrid cloud adoption
So far, data mobility isn’t a solved problem. There are point solutions out there, plus a few companies working to deliver new products that address the major performance and availability issues. However, I still think we’re seeing islands of data being created that need feeding and watering, rather than being self-managing. The issues around data mobility are going to get worse as we move into a more distributed world. There are no standards on how data created at the “edge” is kept consistent with copies on-premises in the core of the data centre and in the cloud. We need some more definitive standards and solutions in 2019. Here are a few related links.
- Conflating Data Protection and Data Mobility
- Data Mobility Technology Choices Series
- Technology Choices for Data Mobility in Hybrid Cloud
- Hybrid Cloud and Data Mobility
- Soundbytes: The Data Fabric Explained with NetApp CTO Mark Bregman
Data Management needs a reboot
In general, data management needs some more effective standards and tools. Look out for a Storage Unpacked podcast (subscribe) towards the end of this week which will cover this. However, in the meantime, here are some other related links.
- #65 – Challenges in Managing Unstructured Data with Shirish Phatak
- #68 – Intelligent Object Storage with Scott Baker
- DataGravity Pointed The Way to Data Rather Than Storage Management
The Architect’s View
The challenges of managing physical storage are today, mostly solved. I say that not in any flippant way, but as a reflection on how the industry is developing. There are lots of great vendors and start-ups bringing faster, cheaper and better products to the market that incrementally address the need for increased performance and reliability. However, move up the stack and we start to see challenges with data rather than physical storage. This will be increasingly so, as we move to hybrid cloud where the physical infrastructure is obfuscated from view and fairly irrelevant.
So, I’m personally hoping for more of a data management focus for 2019 (we’ve started already with this podcast episode) and to see how hybrid cloud really brings value in this area.
Copyright (c) 2007-2019 – Post #5462 – Brookend Ltd, first published on https://www.architecting.it/blog, do not reproduce without permission.