Start Waiting on 3DXP arrays, by Woody Hutsell, AppICU
Let’s get one thing out of the way. Most storage systems will eventually offer 3DXP. Why? Because adding 3DXP SSDs to a storage array will be easy.
A second thing, I think the early usage for 3DXP will flow largely to server vendors (and their suppliers). This is a major point and central to my thoughts on storage and 3DXP. In the server, 3DXP reduces cost and increases density versus RAM.
3DXP in external storage will lag expectations until there are major advances in density and price.
I have worked in the part of the market that 3DXP external storage solutions will target for the last 17 years. For most of those 17 years, I think we could comfortably call this space Tier 0. These are customers whose end-customer satisfaction, missions or revenue are directly tied to the performance of their storage arrays. When I say performance, I really mean latency sensitive. They are so latency sensitive that they will not tolerate storage services getting in the way of application performance. There are customers in the financial, telecom, defense, government, retail, e-commerce and logistics businesses that I could probably with a high degree of accuracy predict their interest in this solution.
These customers are willing to pay for low latency. Customers in this category bought all RAM solid state storage. They were early adopters of all flash arrays. They still buy based on latency curves (who delivers predictable low latency at the IOPS level they require).
These are not the customers buying Tier 1 arrays with a full suite of storage services. They will not tolerate data reduction or storage services if it impacts latency. These are not the customers buying primarily on cost/capacity though they still have budgets and need a solution that fits that budget.
I love this Tier 0 market, because these customers are solving world class problems and must stay on the bleeding edge of technology to grow their business. These customers will buy 3DXP arrays that deliver on the low latency potential of 3DXP. The phrasing of this sentence is no accident, if the array offers 3DXP but only delivers modest latency improvements, it will be largely ignored.
The first enterprise market to hit it big with flash was inside the server, particularly PCI flash (think Fusion-io). The second enterprise market to hit it big with flash, a few years later, was the Tier 0 external storage market (think Texas Memory Systems (subsequently as IBM) and Violin Memory). These splashes were nothing compared to the tsunami of business when all flash arrays entered the Tier 1 market with compelling economics driven by adoption of flash in consumer devices and supported by inline data reduction technologies to further reduce the cost per capacity. These were majority buyers who were confident that the technology wrinkles were ironed out and who by and large wanted better performance than they could get from their disk-based solutions but were very focused on storage services, cost and cost/capacity. They are not Tier 0 buyers though they won’t go back to disk having tasted the sweet nectar of low latency storage.
Tier 1 customers are unlikely to buy into all 3DXP storage arrays until the cost approaches the cost of flash because for these customers the difference between 120 microseconds of latency and 20 microseconds of latency is not as motivating as the difference between 5-20 milliseconds of latency and ½ a millisecond of latency. And can you really get 20 microsecond latency on a Tier 1 device loaded with storage services?
What does this mean for the industry? The market for 3DXP in external storage arrays will appear vibrant due to product introductions but the revenue that can be directly attributed to 3DXP in external storage will be low until the cost and density make meaningful improvements. Storage architects are already designing ways to use 3DXP as a RAM replacement/supplement in the storage array. There is some interesting potential here given the memory requirements for flash metadata and caching and the use of 3DXP as a tier of storage. These steps are reminiscent of the way flash was gradually introduced into Tier 1 before it became Tier 1, for example in RAID cache backups. As with the all flash arrays, the all 3DXP arrays custom built for the best latency curve at the right price will start out in the Tier 0 space waiting for the cost and density improvements that bring it to the big time. This time around, that transition could take much longer than it did with flash based arrays. Flash arrays benefited massively from the density and cost reductions needed in the consumer space. 3DXP does not appear to have the same tailwinds yet.
Woody, the Tire 0 market will move to cloud or not ? I think the DSSD also target on the Tire 0 market and been canceled by Dell EMC.
There are certainly Tier 0 use cases in the cloud. An easy example are software as a service providers whose applications share latency requirements similar to on-premise applications. Similarly, large cloud providers are using Tier 0 for a wide variety of low latency use cases from payment processing to metadata storage. If you are asking, will the need for Tier 0 storage decrease because of applications moving to the cloud, I do not think so. First, I think the mission critical latency sensitive applications that are on-premise today will be the last to move to a cloud. Second, if the application owner is just shifting from on-premise to hybrid or public cloud they will still want the low latency they have been receiving. Finally, I think the rise in real-time analytics such as IoT, web personalization, fraud detection, etc will drive new use cases for Tier 0 in the cloud. I do not look at DSSD being discontinued as an argument against Tier 0’s future but may have been an argument against what they optimized and at what cost.
I agree with your top level conclusion that we haven’t started to ramp demand for SCM arrays today and even when we do begin the ramp it will be quite some time before the revenue reaches the level “All Flash” has reached.
Where are we on that adoption curve? It feels like about 2006 – starting to think about a product like the RamSan-500 rolling out in a year or so. It took flash ~10 years or so to reach double digit percentages of the storage market. Why might this time be faster/slower?
+ Large players that missed participation in the Flash wave don’t want to miss the next big thing and may be willing to invest sooner
+ Perception of solid state devices has changed, they are seen as the more reliable solution. Initial SCM promises even higher levels of reliability
+ The initial gap between SCM and NAND looks smaller than the initial gap between NAND and Magnetic
– The ability of SCM technology to scale has its own challenges – invention required – possibly slower cost decrease
– Performance delta and primary value proposition is not as large as the delta between NAND and Magnetic
– SCM does not solve the same magnitude of problem in the mobile space that NAND did
All that in mind, I think 10 years is as good of an estimate as any for adoption based on price competitiveness with NAND.
I think what’s missing in your line of thinking is could other value propositions of SCM cause an inflection? Could that inflection be sooner? One potential one jumps out – the comparison to memory. For memory use cases the price point is already potentially attractive. Are we headed to an inflection point in External Storage where we adopt SCM and provide a rich set of services around it? Or are we headed to an inflection point in External Memory where we adopt SCM and provide a rich set of services around it? Or better yet an inflection point where External Persistent Memory is adopted? What’s the difference between the 3?
What must be true for some of these other potential futures? What sort of trajectory would that put us on for large scale SCM adoption? Does that change the magnitude of adoption?
Thanks for the interesting blog topic.
Some great observations that expose a strong understanding of the industry and its history. Thank you. Some really hard questions, so let me give it a try.
1. It is like 2006, no credible external storage arrays are offering 3DXP at this point (my apologies to those of you with 3DXP storage arrays in Powerpoint).
2. I think it will take longer for 3DXP to hit double digit percentage market share in external storage than it has for flash. Some reasons why 1) the performance pain in the market has largely been relieved by flash. If the broader market really wanted faster and higher endurance external storage we would still be using SLC. 2) I consider the latency introduced by most external storage arrays that are providing Tier 1 features to be too high even with flash. In other words, the benefits of flash are watered down with high latency storage services. 3DXP will only further expose those inefficiencies. Net, I think this is a Tier 0 market with less total available revenue than the flash market until the prices converge.
3. I like your list of why it might be faster/slower.
4. I think the wildcard for 3DXP will be adoption in the server tier. The server tier is primed to be disrupted by 3DXP. 3DXP is slower than RAM, has less endurance than RAM, but may be good enough. If there is anything we have learned in flash it is that there is a large set of customers that will take good enough over the best. At a minimum, I currently think it will be a major new tier in the server (I have friends in this space who are not quite so sure given the latency penalty of 3DXP vs RAM). It may not cause much of a dent in the amount of RAM put into a server but it may over time take pressure off server vendors to keep offering higher and higher RAM capacities. If I am right, and this becomes a key persistent storage tier in the server then it will actually further slow the external storage use cases for 3DXP. As it relates to the other models, I see rich storage services and 3DXP being at odds with each other due to the latency topic above. I am very much interested in the 3DXP as an external pool of persistent memory concept. Dis-aggregation is counter to hyperconvergence, but I think it is the better of the two models. Having said that, the whole idea of disaggregation requires us to get customers to buy into a new way of doing things and that is much harder than taking an existing technology and making it faster. So, assuming 3DXP takes off and it has more tailwinds than any other new memory technology in the last decade, then I think it will happen first in servers and second in Tier 0 storage (by which I mean solutions like the FlashSystem 900) that can really be optimized to deliver on the low latency of 3DXP. The biggest risk and it a big one, is that since this market is expected to be driven by enterprise use cases it will not have the monumental benefit flash did being driven by mobile use cases and thus the density and price may not improve at the rate that will be needed to fuel real growth. For what it is worth, as someone who loves low latency, I really hope I am off on this and that customers benefit sooner than later from lower latency storage in Tier 0 and Tier 1.
[…] 1:46 pm and is filed under Marketing, SSD. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own […]