In our annual yearend podcast and it’s the Ray and Howard show, talking about storage futures, industry trends and some storage world excitement of- the past year.
We start the discussion deconstructing recent reductions in year over year revenues at major storage vendors. It seems with the advent of all flash arrays (AFA), and all major vendors and most startups now have AFAs, customers no longer feel the need to refresh old storage hardware with similarly (over-)configured new systems. Instead, most can get by with AFA storage, at smaller capacities that provides the same, if not better, performance. Further9, the fact that AFAs are available from so many vendors and startups, customers no longer have to buy performance storage exclusively from major vendors anymore. This is leading to a decline in major vendor storage revenues, which should play itself out over the next 1-2 years as most enterprise storage systems are refreshed.
Recent and future acquisitions also came up for discussion. NetApp’s purchase of SolidFire was a surprise, but SolidFire had carved out a good business with service providers and web-scale customers which should broaden NetApp’s portfolio. In the mean time, the Dell-EMC acquisition takes them out of the competition for new technology acquisitions, at least until it closes. NetApp’s new CEO, George Kurian, appears more willing than his predecessor to go after good storage technology, wherever it comes from.
Software delivered (defined) storage came up as well. With the compute available in todays micro-processors, there’s very little a software delivered storage system can’t do. And with scale-out storage, there’s even more cores to work with. Software delivered storage and scale-out will continue to play a spoiler role, at least in the low to mid-range, in the storage market throughout the next year.
Nonetheless, hardware still has some excitement left. Intel’s recent acquisition of Altera, now makes Xeon/x86 processing available for embedded applications that previously had to rely on ARM and MIPS processing. Now, there’s nothing an FPGA hardware based system can’t do. Look for lot’s more activity here over the long term.
We talked about recent SMR disks coming out and how they could be used in storage systems today. There was some adjacent discussion on the flash-disk crossover, and conclude it’s unlikely over the next 3-5 years, at least for capacity drives. Although there’s plenty of analyst that say it’s already happened, on a pure $/GB there’s still no comparison.
We then turned to 3D TLC NAND and the reliability capabilities available from current controlller technologies. Raw planar NAND available today is much less reliable than what we had 1-2 generations back, but the drives, if anything, have gotten more reliable. This is due to the reliability technology inherent in todays SSD controllers.
We had an aside, on SSD overprovisioning and how this should become a customer level option. Reducing overprovisioning would decrease drive endurance but it’s a tradeoff that the vendors/distributors make for customers today. We feel that at least for some customers, they could make this decision just as well. Especially if drive replacements were a customer maintenance activity with replacement SSDs shipped in a just-in-time manner.
We conclude on 3D XPoint (3DX) non-volatile memory. We both agreed 3DX adoption depends on pricing which will change over time. In the long term, we see the potential for a new storage system with 3DX or other new non-volatile memory as a top performing storage/caching/non-volatile memory tier, 3D TLC NAND as a middle tier and SMR disk as the bottom tier. When is another question.
Our year end discussion always wanders a bit, from high end business trends to in the weeds technologies and everything in-between. This one is no exception and runs over 49 minutes. We tried to do another Year End video this time but neither of our video recording systems worked out, but we had a good audio recording, so we went with the podcast this year. Next year should be back to video. Listen to the podcast to learn more.
Podcast: Play in new window | Download (Duration: 49:42 — 22.8MB) | Embed
Subscribe: Apple Podcasts | Google Podcasts | Spotify | Stitcher | Email | RSS
Howard Marks
Howard Marks is the Founder and Chief Scientist of DeepStorage, a prominent blogger at Deep Storage Blog and can be found on twitter @DeepStorageNet.
Ray Lucchesi
Ray Lucchesi is the President and Founder of Silverton Consulting, a prominent blogger at RayOnStorage.com, and can be found on twitter @RayLucchesi.
Great stuff, Howard, Ray.
I’d add my two cents from what we see from our corner of the market:
1) All-flash/SSD is becoming a commodity – in the sense of production deployments are now expected to deliver “all-SSD” level of performance.
2) SDS is a paradigm shift – people want to move away from proprietary specialized storage boxes – be them traditional SAN or all-flash arrays. The drivers are simplification – getting rid of specilized single purpose hardware; reducing lock-in (in reality “lock-in free” is a wish, but hey – better be locked on the SW, rather than SW+HW) and of course slashing cost (this is really IT Stack TCO, not just storage).
3) New technologies – SMR, 3DX, ARM, etc. are all welcome and will fuel further storage solutions/tiers, however they are subordinate to the shift to SDS – it’s the software that will be leading and will squeeze every I/O and MB/s out of the new hardware. Not that “SDS” is something new as a concept, but as you know best – the industry is rather cyclical and now this approach is on the rise. Initially as you say “in the low to mid-range, in the storage market”, although we see this cuts through the tiers and now goes into mid-to-high-end systems as well.
We’ve also summed the trends we expect to develop/accelerate in 2016 on our blog (https://storpool.com/blog/storage-trends-and-predictions-2016) and they come almost 1:1 with yours.
Cheers,
Boyan
Boyan,
Thanks for your comment. Yes SDS is still a hot topic, how far it reaches up market is an open question. Although, there’s good adoption in service providers and SMB-Mid-range data centers. It’s still an open question whether enterprise class customers want to depend on it. Nonetheless, it’s interesting to see where things are headed.
Ray
Ray,
Sure, as any other technology – it has places where it fits best and others where it’s not a good fit. Big enterprise customers will probably take a couple more years before they adopt it.
It’s also interesting to see the new crop of 2016 vendors that will make products with 3DX and SMR (mostly). They’ll maybe take more time to become new tiers of storage, but for example we see good potential for SMR becoming the backup tier and finally killing tape and bringing “cold data” (near) online. Let’s see.
Boyan