The future of data storage

Following on from the previous post, I posed this question over on NPI…

http://discussion.fool.com/future-potential-in-data-storage-3237…

Given you’re a sizeable bunch of clever growth investors with a lot of technology knowledge, I’d be really interested in thoughts around here on this too…

Thanks
Ant

(Posting copy…)
Guys - if there’s any smart ones amongst you who know about data storage I’d be interested in your views.

I had invested originally in EMC and NetApp and did very well during the 50% growth era in the 2000s however as mixed flash and then all flash started to emerge things just went sideways.

I turned to investing in the flash players and did well with Fusion IO (or whatever it was called) and Nimble and got nowhere with Violin.

I remained in EMC since I really liked the sum of their parts with VMware plus the potential in their Pivotal business and believed they and NetApp would eventually migrate to all flash and dominate in that too. Given data traffic etc is going up 100x through to 2020 I naturally thought storage demand would keep going.

In recent quarters it feels as though All flash has really arrived with Pure and Nimble doing well. Together with the rearranging of the furniture by Dell’s buy out of EMC it feels that the landscape is really getting transformed.

Does anyone have any thoughts on the best opportunities in this space and how to play it?

I would have preferred a Pivotal focused EMC and I think Dell is getting it all on the cheap which together with their issuance of special VMware tracking stock looks really messy.

Pure is still not public.
Nimble is doing ok.
NetApp just isn’t getting it together.
Violin lost the plot.
DELL etc are not really pure play storage.

Any ideas on whether this is an attractive new paradigm opportunity moment or is there nothing to see here and we should move along?

Ant

2 Likes

I know its bad form to reply to ones own post but I need to make a correction…

Ok I must have been asleep at the wheel on this as Pure has now listed and has halved from its peak and since I sold out of Nimble at what turns out the peak at 27 it has fallen to $8 which now looks accessible.

I guess these 2 could be the immediate candidates to consider.

Ant

The chief threat to the storage vendors is Amazon and Google. They are building massive storage service business and they do not buy their gear from the Storage vendors. They cable together their own storage farms from commodity disks and servers. The storage vendors do not make a dime.

4 Likes

they do not buy their gear from the Storage vendors.

Not entirely true. Both AWS and google service enterprise customers with specific SLA’s and in those segments they use many storage vendor solutions. I know this for a fact.

The chief threat to the storage vendors is Amazon and Google. They are building massive storage service business and they do not buy their gear from the Storage vendors. They cable together their own storage farms from commodity disks and servers. The storage vendors do not make a dime.
<<

On the surface, this may be true. But if you understand what the storage vendors are actually doing – at least the ones that will survive – the landscape is very different.

For disclosure purposes, I am a NetApp employee (for a long, long time actually), so take that fully into account here. I will avoid speaking specifically about NTAP or its competitors’ offerings except where it is super-relevant to your statement above.

Flash storage has changed the landscape for everyone. Western Digital, Seagate were the largest HDD manufacturers, and they’re not doing as well. They sold drives to the major storage vendors (EMC, NTAP, Dell, NMBL, PSTG, etc.)… until they couldn’t. The small startup vendors that came out with flash-based point products have done ok… until they didn’t. See NMBL and PSTG stock price for confirmation. Point products are fine until your need for features/flexibility outgrow the glossy brochure or raw speed of just having flash storage in-house.

The larger storage vendors – NTAP, EMC, IBM, and to some extent, HPE still – all bought disks from WD, Seagate, Hitachi, etc… until the cost/GB for flash became attainable. Some of those vendors, including my employer, did a great job incorporating flash into their products, and some didn’t.

Two or three years ago, those companies that looked (or are looking) at flash only as the “next thing we have to be offering” were probably already doomed. Cloud-based storage offerings began being more compelling, and interweb speeds fast enough to make cloud storage actually usable for the masses. So those companies that embraced the cloud, or a “hybrid cloud” offering were probably on the right path.

Storage vendors that have cloud or hybrid-cloud offerings are absolutely making money – and quite a lot of it. The vast majority of users that opt for cloud storage still need to keep some things in-house, and it makes a boatload of business sense to be able to manage all of that storage together – for efficiency, cost, employee training and many other reasons.

Again, I have some built-in bias, so take that into account…

IBM, in my opinion, is a dead shell of a company. Might continue to dole out big dividends, but they’re eating their own and not innovating much in terms of storage, so you can mostly ignore them in this realm.

The EMC/Dell merger finalizes next week, and there’s so much FUD from that, that all the other storage vendors will benefit for awhile as customers run to safer territory. And the new Dell will suffer from talent drain as they go private and have to tighten the belt even further, so mid-term there’s even some viability concern for R&D.

Nimble is a point player. If you want fast iSCSI only, great. But they’ve never made a profit, and the market has evaluated the stock price accordingly.

Pure Storage is also a point player. Probably IPO’d too early, and won’t be able to show enough to keep the price up.

NetApp is one of the few storage vendors that looked past just offering fast, local storage when it was sexy. While the stock price floundered, the recent jump seems to have correctly accounted for the “ok, you do know what you’re doing in this changing landscape” factor. No guarantees, and no insider info here, just my $.02, but having the most widely-deployed file system in the world available for use in-house, on fast flash arrays, in the cloud, or on ALL of those simultaneously, is a pretty compelling argument to me as a shareholder, and customers – the ones writing the checks – seem to agree.

17 Likes

(Disclosure: I work in this industry…)

The flash players (point products) will serve their natural purposes, but most will die off or be acquired. Flash is not the “end state” for most storage, as I see it.

For the majority of customers requiring storage (not including secure-type governmental agencies or financials where they may have laws requiring them to have in-house storage), their data is moving, either in part or in majority, into the cloud. This is both a cost savings from admin overhead and internal storage costs, as well as from offloading to a third-party that can absorb the overhead across a larger scale.

So if you want to invest in storage, you have to be looking at vendors/companies that are poised to take advantage of the cloud or hybrid-cloud ecosystems. Vendors that have well-known, scaleable storage systems and file systems that can operate seamlessly across all of that (local [spinners + flash], cloud-only, and hybrid-local+cloud) storage.

Working in the industry myself, I can tell you that there are only maybe 3 companies that will stay relevant in the move to the cloud. Pure, Nimble, Violin, Nutanix and others like them are not likely to land in that list of three, nor is IBM.

Dell (which will no longer be EMC after next week) is still a question mark on the cloud/hybrid-cloud part, and they will have their hands very full figuring out how to be a private company and deal with the inevitable brain-drain of talent and the uncertainty from customers. The VMware tracking stock mess just muddies the water even further.

HPE still has some clout, but they just split off from the “other” HP, so there’s some small amount of risk there. But they’re somewhat viable.

NTAP has the most widely-deployed file system in the world and a highly-scaleable clustered storage system (tens of nodes, each with so many terabytes of storage it would scare you), which can run on local, cloud or hybrid-cloud environments. And can all be managed seamlessly across all of those environments, and move data in and out of the cloud easily.

Choose your poison, but IMO it wouldn’t be any of the point players that will die out.

8 Likes

HPE still has some clout, but they just split off from the “other” HP, so there’s some small amount of risk there. But they’re somewhat viable.

Full disclosure, I’m retired from HP but not in any way that influences my opinions on this technology. I was in the part that was EDS (before being bought by HP), is now HPE and soon to be spun off to merge with CSC. So, nothing to do with hardware at all. I don’t own any HP or HPE stock (although this is causing me to consider HPE).

This discussion brought to mind something I was investigating recently. HPE is supposed to debut “the Machine” soon. It’s billed as “the future of computing”. That’s a pretty bold statement, but I wonder it they are on to something. So, how does that relate to this data storage discussion? Well, the most exciting part of “the Machine” to me is the memristor technology, which is memory, cache and storage all in one.

If we would have a single technology that would have the speed of cache while having the persistence of storage, we could combine storage, memory and cache into a single device that would keep all the information and instructions. There would be no need any more to boot-up computers, to shut them down, or to hibernate. All instructions and data would just be there when we needed it.

Using the memristor technology, HP prototyped a crossbar latch memory that can fit 100 gigabits in a square centimeter, and proposed a scalable 3D design (consisting of up to 1000 layers or 1 petabit per cm3). In 2012, the device achieved a read time of 90 nanoseconds (if not faster), approximately one-hundred times faster than contemporaneous flash memory, while using one percent as much energy.

This eliminates the need for separate storage devices (except maybe backups?). A big data farm could be just an array of these “machines” that combine a process and storage on a single card. I can’t find it now, but I think I read earlier that a single card could have multiple terabytes of memristor storage. I use the term “storage” but it really is different. It’s more like memory (DRAM) because it’s accessed directly by the processor. There is no longer any need to fetch data from disk/flash memory like is done now. It really does look like a possible game changer to me, provided it’s not bungled by the company management.

Here’s a link to the site if you want to read more.
http://www.labs.hpe.com/research/themachine/

Steve

6 Likes

So Hlygrail and Sagewren - one question. Just who are the cloud hosts buying their storage from. I assume they do not make their own so it has to come from somewhere right? Just because NTAP, EMC, DELL and HP are not selling to individual companies anymore presumably the same amount of storage is needed just in the cloud and their customer base becomes the cloud intermediaries.

So AWS, Msoft Azure, Google and all the cloud operators like BOX and the cloud datacenters like Digital Bay must be the customers buying for cloud storage on behalf of end customers.

Why aren’t the usual suspects or newcomers doing just as fine selling to the cloud operators rather than end customers?

How is their a famine for growth amongst the storage purveyors if we are to presume that the cloud players aren’t making their own, nor buying at source (Western Digital or Sandisk etc) or from new point solution providers of flash.

Someone has to be making storage to shove into the cloud - isn’t just thin air.

Ant

1 Like

Ant,

I am not in the industry, but am an info sponge. My understanding is that AWS and Google at least, and I’m sure the rest to stay competitive, have developed in-house software that enables them to use off the shelf components and commoditize the hardware, at least as much as they reasonably can.

Data storage continues to grow, as does bandwidth, but it is software that takes the money in, not hardware boxes. Just like it is for bandwidth.

These in-house options are what allow the scaling that you see with the top Cloud players.

Not that anyone has anyone one cover all solution but it is my understanding that this has been the trend, and therefore why the once mighty storage giants are not doing so well with the advent of the cloud.

Tinker

1 Like

Ant, most of the cloud storage is plain old disk, not the fancy networked cached stuff that the EMC’s of the world are selling.

The larger storage vendors – NTAP, EMC, IBM, and to some extent, HPE still

I think HPE has higher market share on storage than IBM.

I think HPE has higher market share on storage than IBM.

You’re correct, and I didn’t list them in any storage market order. The “to some extent” part about HPE was related to the fact that (like Dell/EMC will undergo today) there is some uncertainty in the mergers and spin-offs actions that could impact their market share.

And to answer/address a previous question…

Just who are the cloud hosts buying their storage from. I assume they do not make their own so it has to come from somewhere right? Just because NTAP, EMC, DELL and HP are not selling to individual companies anymore presumably the same amount of storage is needed just in the cloud and their customer base becomes the cloud intermediaries.
<<

AWS, Google, Microsoft Azure… they’re all still largely running their clouds on spinning disks (“spinners” as they’re known in the industry, to differentiate from flash). They have built a boatload of software to manage across thousands and thousands of white-box PCs to make it look like a giant pool of storage, that they can then carve out and assign to a given customer. That same software handles hierarchical storage processes (putting the least-used stuff on slower disk or less-expensive tape even) and redundancy to address inevitable disk failures.

I have to note that the spinning disks used by large storage vendors (NetApp, Dell, HP, Hitachi, etc.) are not the same hard drives that you get at Best Buy. In almost all cases, they have firmware customized heavily for specific storage systems or file system behavior, and I’m sure in at least some cases they also specify improved MTBF (mean time between failure) and other things in their contracts with the spinning-disk manufacturers – mainly Seagate, Hitachi and Western Digital, with Samsung still climbing.

So, to answer, the cloud companies are pretty much buying consumer-grade disks directly from the disk manufacturers themselves, putting them in the cheapest generic PCs they can buy (more or less) and then using software to manage clusters of thousands of “nodes.” They never really bought much from the large storage vendors anyway, so the direct impact didn’t change. What has changed some is that end customers are moving some of their own storage to the cloud.

The compelling argument for cloud storage is cost – rather than buying $xxxK of storage to put in your own data center, you pay someone else to store it for you, for less $$ than you can do it yourself. The cloud companies can amortize their overhead across many customers. But cloud storage is not everything, can’t handle everything, and shouldn’t handle many things…

I’ll refrain from commenting on which storage vendors are doing a good job innovating toward cloud and hybrid-cloud storage, and which ones are mostly FUD… you can do your own research and make your own decisions.

4 Likes

HPE is supposed to debut “the Machine” soon.
http://www.labs.hpe.com/research/themachine/

Some interesting technology, for sure. I have always respected certain parts of HP’s business – the LaserJet line of printers is still the gold standard in some ways (I’ve been printing on a LaserJet 6P that I got for $30 at a state surplus auction for probably 10 years now), for example.

Back before they killed off Compaq, I had a chance to visit their R&D labs in Houston. This would have been circa 1994 or 1995. We signed a bunch of NDA paperwork, and then they showed us where they were on resiliency for memory failures – you could pull out an entire stick of failed DRAM with the server still running, and their “RAID for memory” would keep things running while you swapped out the stick of RAM. It was pretty impressive back then.

Just as a comparative along the same lines of “the machine”… I recently built a new computer to replace my 6-yr old PC. The “old” PC is still very, very usable because it was built with the highest-end stuff I could buy six years ago. It has ~16TB of spinning disks in it, and a whopping 8GB of (not enough) memory.

The “new” computer started its life with a 512GB NVMe flash “drive” that cost ~$200. But what is notable about that is that this 512GB of storage was smaller than a single stick of DDR4 memory that went into the same box. And obviously 10x faster than any spinning disk, too.

Check out the picture below with one of them next to a US quarter:

http://www.storagereview.com/images/Samsung-M_2%20NVMe%20SSD…

32 of these would cover the ~16TB of spinners in the old PC, and take up about the physical space of half of a 3.5" disk drive. Still far more expensive, but I was still impressed to hold 512GB in the palm of my hand, with the weight being roughly the same as a piece of paper!

2 Likes

The “to some extent” part about HPE was related to the fact that (like Dell/EMC will undergo today) there is some uncertainty in the mergers and spin-offs actions that could impact their market share.

Thanks for the insight. Appreciate it. Just a quick comment on HPE. HPE (post split enterprise company) has 3 major divisions Enterprise Group (EG), Enterprise Services (ES) and Software. My understanding is these 3 divisions more or less act as separate entities. The Enterprise Group comprises of Server, Storage and Network. The ES will be spin-off and merged with CSC. By and large this doesn’t impact EG and its sales or revenue’s impact is limited to the extend that ES will not necessarily position EG solutions as the first choice. I also believe as part of the separate the ES will continue to sell EG solutions.

So I am not expecting any immediate change in market share for HPE due to merger, spin-off, etc.

Just to complete the story, HPE is also planning to sell its software division there are bids for $7.5 B and HPE is looking for $8 to $10 B. At the end of all this HPE will be left only with EG, which is a pure Server, storage, network company and about $20 ~ $23 B in cash. Considering the market cap just about that at sometime back this year, it is pretty amazing.

1 Like

Regarding HPE, Bert’s got it all right here:

http://seekingalpha.com/article/4004190-hpe-will-thoma-bravo…

HPE is also planning to sell its software division there are bids for $7.5 B and HPE is looking for $8 to $10 B.

The spin-off is done deal now. Software division will be spin-off and merged with microfocus and HPE shareholders will get 50.1% of new company and HPE will get $2.5 B cash, valuing the SW division @ $8.5 B.

http://www.cnbc.com/2016/09/07/hpe-to-spin-and-merge-non-cor…

…and the Dell/EMC merger is also a done deal. EMC no longer exists, and we’ll have no more public info on their storage market share in terms of shareholder-reported earnings.

That, and EMC-now-Dell will continue to bleed staff, brainpower and leadership (it’s already happening, and I am VERY close to a now-ex-insider who confirmed my suspicions). Dell will HAVE to reduce their expenses something fierce to be able to justify and pay off such a monstrous debt (64 billion might as well be an infinite number to me) in any reasonable timeframe, and who wants to be on that ship as the clouds keep getting lower?