CXL: A Basic Tutorial Flash in the Era of Digital Infrastructure: What CIOs Need to Know
Guest Post

Evolution of SSD Storage in Public Cloud

IDC analyst Andrew Smith explores flash-based services in the cloud. Review key offerings and how cloud providers are delivering services to the market and using flash and SSD to bring about those services.

00:19 Andrew Smith: Welcome, everybody to the "Evolution of SSD Storage in the Public Cloud." My name is Andrew Smith. I'm an analyst on IDC's infrastructure team focused on public cloud storage services and some of the associated secondary workloads, but today we're going to go through and just kind of a recent history of flash-based services in the cloud, some of the key offerings and how cloud providers are delivering services to the market, and using flash and SSD underlying those services.

01:00 AS: I want to give you a couple of data points to help quantify the growth of SSD in the cloud. It's one thing to say anecdotally that we know that the use of flash is growing, but it's another to provide some empirical data points to help you triangulate what growth actually looks like. So, we'll go through that. The next I want to talk about opportunities and benefits of SSD in the public cloud from two perspectives -- one from the cloud service provider perspective and from the customer or the end-user perspective -- because I think that they both have some key benefits that should be highlighted independent of one another. And then, finally, some concluding remarks and key takeaways. I'll also touch base on kind of our . . . a little bit longer term outlook of this market, how we expect to see continued growth, especially in the context of COVID-19, and how we think that will impact the use of SSD in the cloud overall.

02:08 AS: So, let's start with a brief overview of services, slide four. This is a quick timeline of those select service and their availability that I was referring to. So, really, SSD-based services and flash in the cloud are pretty recent history. Starting in 2012 is when AWS added an SSD option to EBS, but it was really kind of considered a premium option or a premium storage tier at that point. 2014, Google added an SSD option. 2017, Azure added an SSD option with managed disks. Oracle in 2018 with OCI Block Volumes. And throughout that timeline, we've also seen Google, AWS add . . . continue to add different flavors or tiers of flash to some of their services.

03:10 AS: Providers who aren't on here, who are important to note, IBM has had a flash-based block tier for several years. 2019 we saw the release of Azure NetApp Files, which is really a great example of a storage OEM partnering with a cloud services provider to bring a solution to market.

Really, why I laid these services out on a timeline is to just show how leading providers have continued to add SSD and flash-based options to their portfolio over time, and it's been through a mix of organic product development as well as acquisition. I think it's indicative of the opportunity that they see in this segment, and from a customer's perspective, what's important to point out here is flash-based storage services have really moved from what was considered more of a premium option in 2012 to, in 2019 to 2020, they're really a baseline offering. If you're looking for a block- or a file-based storage, cloud storage service, many providers are now offering an infrastructure that is SSD- or flash-based as that primary tier. It no longer needs to be denoted by customers that they want to use that high-performance tier, so really indicative of how these services have matured just in the past few years -- this slide -- so we've seen.

04:51 AS: I've shown you from a vendor, from a provider perspective, how they have evolved their portfolio over time, what about from the workload perspective? So, this is end-user data from our, IDC's IaaSView survey, and we've asked the same question over the past three years. We ask what deployment scenarios are being used along with public cloud IaaS . . . that I want to highlight is, as we've asked this question over the past three years, that top choice used to run production or external-facing applications has slowly crept up the list over time, really indicative of how end-user cloud deployments are maturing and they're relying on those services -- IaaS services -- for a business-critical or primary workloads.

Further down the list, you see things like used for data archive, used for backup and disaster recovery, those are kind of what we would consider some of the entry-level workloads associated with cloud IaaS usually based on object storage, those have slowly made their way down the list. As like I said, customers are maturing their usage of IaaS; they're moving towards those higher performance, more primary business-critical workloads, and that's where flash becomes increasingly necessary, increasingly critical to running those workloads at the performance, the accessibility, the availability and the SLAs that they're required.

06:37 AS: So, again, this is just a way of looking at it from the workload perspective, how customers are more reliant on those services. Some of my colleagues are going to focus much more on the workload aspects of these segments, so I'll leave that to them, but this is just to give you a little bit more context.

06:55 AS: So, just kind of summarizing this section, again, flash and SSD-based storage services in the cloud have really gone from a premium-tier choice that a customer made to a standard tier that the customer is provisioned upon their initial request. It's driven by that growth of production applications, reliant on public cloud IaaS as well as some emerging use cases which are really driving the proximity of high-performance storage to compute things like analytics, database applications, some of those more emerging use cases. And as public cloud IaaS adoption grows, we think customers will increasingly expect performance storage services which meet or exceed on-premises alternatives. It's all about delivering that same customer experience that customers have come to expect using an on-premise storage environment.

08:01 AS: And at the end of the day, it's about the SLAs and the performance. A customer might not even be aware or care that the service, the storage service that they're buying is SSD-based or flash-based. They're looking for that performance to meet an application need perhaps or a workload need. And as long as those SLAs are met, in many ways they're happy, so that's another consideration to keep in mind when you're thinking about this from the workload perspective. So, we've talked about the services being deployed, some of the workload requirements of customers, a lot of anecdotal data, but I want to give you a couple empirical data points to show what we're seeing in terms of the growth of flash to in the cloud.

08:57 AS: So, the first data point is from IDC's Cloud Infrastructure Index, and really the CII, the Cloud Infrastructure Index, tracks demand from infrastructure . . . Excuse me, tracks demand for infrastructure from top public cloud providers, and this is giving you a two-year growth snapshot from 2016 to 2018, HDD capacity on the left, SSD capacity on the right, and you can see that SSD capacity over that two-year time span grew 200%, so it's almost double the rate of capacity growth of HDD. And this just kind of gives you an idea of just how much capacity those leading public cloud services providers are consuming.

09:54 AS: So, this is, I think is one of the . . . A first good data point to use to quantify that demand. The second data point I want to run you through is from IDC's ODM Direct research, so our ODM Direct segment tracks revenue for hyperscale infrastructure, both service and storage sold by original design manufacturers or ODMs directly to cloud services providers. So, this is actually emerging data that we're still working on releasing, but what we . . . From our initial research, what we're seeing is, if you look at 2019 and you split ODM Direct revenue by media type, about 9% of that revenue can be allocated to the sale of SSD-based infrastructure to cloud services providers.

Now, we obviously don't have full-year 2020 data to compare that to give a true annual compare, but if we look at the data we're finishing now for the first half of 2020, we see that that percentage of SSD shipments has expanded to 11%. So, a couple of percentage points of growth. When compared to the total of 2019, it's indicative of growth either way, and I think if you were to play out this growth over 2020 to give a true annual comparison, you'd see that 11% expand potentially one or two more percentage points.

11:44 AS: So, really, again, this is from a slightly different perspective, this is from the ODM perspective, but also it continues to show how revenue associated with SSD and flash sales to public cloud providers is expanding. And, finally, the last data point is from IDC's Public Cloud Services Tracker data.

12:07 AS: So, this is . . . What we're looking at here is services revenue. So, this is the revenue that AWS, Azure are earning on their storage services, and this includes . . . This is a share of object file and block of that total revenue per year out over the past five years. So, what I really want to draw your attention to is object makes up the majority of the market, but the blue and orange segments, looking at how that share has expanded over time, is the key indicator here. Those file- and block-based storage services that are reliant on that underlying flash infrastructure have grown significantly over time and continue to make up a greater share of this market. And I'll give you . . . just to kind of put this into a little bit better perspective because these are just share . . . percentage shares of the total revenue.

13:14 AS: So, in 2019, if you take the file and object segments that equates to $8 billion in revenue, and the annual growth of the file segment was over 200%, and the annual growth of the block segment was over 50%. So, really, these are segments which are making up an increasing share of the services, the storage services market, and they're growing at significant rates, and they're enabled by flash and SSD storage as within the underlying infrastructure. So, this gives you a little bit of the kind of end-user or services provider perspective of how the market is growing.

So, to summarize, the three data points we've gone through overall illustrate the growing use of flash and SSD storage by public cloud services providers. Data point one was capacity consumed by cloud infrastructure providers. Data point two was hyperscaler infrastructure sold to cloud services providers, and then data point three was the storage services revenue earned by cloud services providers. The story for all three data points -- whether it is from the consumer perspective, the cloud services provider perspective or the hardware supplier perspective -- is growth.

14:48 AS: So, hopefully that helps you put together a little bit of a story empirically about growth across each three of those perspectives and paints a positive story overall of the use of SSD for cloud storage services.

So, now let's shift gears a little bit and talk about the opportunities and benefits of SSD and flash in the public cloud from both the customer perspective and the services, the cloud services provider perspective because I think there's a couple of different ways to look at this. So, the best way to think about the opportunity overall is to think about storage accessibility and performance requirements in terms of hot, warm or cold.

15:45 AS: So, the back-of-the-envelope calculations, the rule of thumb that we usually give is that 10% of enterprise data is hot, 30% is warm and 60% is cool or cold. This can vary depending on your definitions of access, but for the purposes of this presentation, we'll say that hot is data that's accessed . . . Hot or warm is data that's accessed either hourly, daily or weekly, and everything else is cold. So, data that's accessed monthly or less frequently, we would put into the cold category. The key opportunity for SSD- and flash-based storage in the cloud is to address and enable that 40% of workloads or use cases which are considered hot or warm and require that kind of profile of fast access to compute and redundancy availability.

16:44 AS: And one thing to point is, we don't expect these ratios to shifting significantly, we don't expect hot capacity to make up 15%. These are pretty standard segmentation that we have not seen move, but the important point to make is that all enterprise data is growing. So, we typically see, on average, enterprises say they expect their storage to grow 30% annually, so really that expanding enterprise data is boosting the capacity of all three of these segments, and what that will lead to is . . . The question becomes, "Well, what types of solutions will meet the needs of these workloads. Will it be a mix of on-premise and cloud solutions? How fast are enterprises shifting and adopting those types of solutions from cloud services to provide that kind of on-premise experience for hot and warm workloads?" I think that's really the interesting question, but this gives you a way . . . I think that 40% of warm and hot workloads to think about conceptualizing, where is the sweet spot, thinking about storage on a spectrum from hot to cold, and what the opportunity is. Now, the same spectrum is important from a customer perspective in terms of their priorities and considerations of these workloads. So, customer requirements for storage services changed pretty dramatically across the spectrum.

18:26 AS: For hot and warm services, the cost of storage is a factor, but it's more about . . . It's not the primary consideration, it's more about the performance and availability and accessibility of that storage, whereas when you get down to cold storage services, really the long-term cost is what drives that performance storage tier and customer considerations to adopt services, cloud-based services in that space. So, in some ways, this spectrum from a customer perspective becomes more stratified over time where cold . . . On the cold side, customers are looking for cloud storage services that deliver that lowest possible cost. Warm hot services customers are dependent on finding services that provide high performance. So, increasingly, it's going to be about the media types that enable this, and on the cold side, it's a mix of HDD optical tape that enable that. On the warm and hot side, it's really left to flash storage to help customers achieve the SLAs and the performance they need. So, I think you'll see this increasing stratification along the spectrum in terms of the media used.

19:55 AS: Now, I guess one caveat is, on the cold side we do see some SSD used on the object tiers, but they're really . . . It's for caching mechanisms at the infrastructure level. It's not something that a customer really has any knowledge or consideration of, and it doesn't really move the needle, so to speak, in terms of suppliers delivering solutions to cloud services providers. So, really, in terms of media, how it's being adopted, how it's being used from a customer perspective, we'll continue to see SSD and flash within that 40% of warm and hot as the core enabler of those workloads for customers.

20:39 AS: So, in summary, this is just a wrap-up slide to give you some of the key benefits from both the cloud services provider as well as the customer regarding the use of flash and SSD. From the cloud services provider perspective, it's about that providing the highest performance, highest IOPS within their portfolio, connecting high-performance compute with storage.

21:07 AS: Also, one thing we haven't touched on is from a data center perspective, right? The high density of SSD storage compared to HDD can typically be a benefit for cloud services providers in terms of saving space within their data center footprint. From the customer perspective, again, customers wanting to deploy more primary workloads within the cloud, leveraging flash storage to keep that high performance, high availability of storage which they require.

And a second thing to consider is just how customers are increasingly relying on the abstraction of storage management tasks that are provided by cloud services providers, so they can focus on the workload, the application, and all of that is enabled by cloud services providers, as well as the performance that flash provides.

22:13 AS: Some concluding remarks and key takeaways just to wrap all of this up. Let's start with some future outlook and some of the considerations that go into our long-term thinking around the performance of flash and SSD in the cloud.

We touched on this a little bit in the previous slide, but one important thing I want to bring up again is price reduction, especially from a cloud services provider perspective, we saw some significant price erosion in 2019, but we don't really expect this to continue and there's a lot of detail on this in IDC's Worldwide Solid State Drive Forecast, which is conducted by some of my colleagues. But what I want to point out is that our expectation is the blended overall SSD price per gigabyte is expected to decline at a more normalized rate, a compound annual growth rate of 13.5%, which we think is reflective of the current supply growth expectations and the outlook for the market overall. I think the point is that 2019 was a bit of an anomaly in terms of price reduction of flash storage -- we expect it to normalize over the forecast period.

23:49 AS: From a provider, a cloud services provider perspective, this is important. They're looking for consistent measurable price reductions over time, so they can forecast their cost of services, their operating cost of services. They rely on that kind of understanding those trends over time, point to be an important indicator to cloud services providers to continue purchasing and enabling those services over the entire forecast period.

24:22 AS: The second consideration I want to focus on is the impact of COVID-19. So, overall, despite COVID-19, we expect the SSD forecast to remain strong and healthy. Though kind of the length of the pandemic, I think is one of the key unknowns variables that will continue to impact the forecast significantly, but from what we've seen from 2020 so far, the SSD market has navigated the market uncertainty well. We've seen continued strong demand from cloud services providers in terms of purchasing and adopting, and that has played through all the way to . . . We've continued to see, from a revenue standpoint, stronger earnings from cloud services providers, certainly no indications of a pullback of cloud services; in fact, it's been slightly counterintuitive. We've seen, if anything, an acceleration of adoption of cloud IaaS as customers are using COVID as a way to accelerate their transformations, their digital transformation, in any way that they can due to heavily, more heavy reliance on remote work in some cases, and just a desire on the point from an enterprise perspective, a desire to modernize their infrastructure wherever possible.

26:04 AS: So, let's wrap up with some key takeaways. The first, going back to that timeline slide, is that among all leading cloud services providers, SSD and flash storage has moved from being kind of an exotic or premium storage tier to entry level to a standard part of many block and or most block and file offerings, just kind of highlights the growing adoption from both a cloud service provider and from an end-user perspective.

The second key takeaway goes back to some of the data points, the growth of high-performance block- and file-based workloads in the public cloud will continue to drive SSD adoption by cloud services providers, and we do not expect any significant slowdown due to current events, namely COVID-19.

26:58 AS: And the third key takeaway is really over the long term, we expect buyers to continue to rely on high-performance SSD and flash-based storage to meet their requirements for hot and warm storage workloads or storage use cases. So, again, back to the stratification of that spectrum, in terms of media, HDD tape optical are going to be relegated to that cold segment where SSD, flash media is going to be the media of choice for that hot/warm 40% of enterprise storage. Again, highlighting that key opportunity. So, with that, we'll wrap up and move on to Q&A, and thank you for your time. Look forward to answering any questions that you have.

Dig Deeper on Flash memory and storage

Disaster Recovery
Data Backup
Data Center
Sustainability and ESG