The Final Post: Hardware Is Dead

Hanging up the jersey

Well, my friends, this is it. The time has come to retire the flashdba jersey after more than seven years of fun and frolics. In part one of this post, I looked back at my time in the All-Flash storage industry and marvelled at the crazy, Game of Thrones-style chaos that saw so many companies arrive, fight, merge, split up and burn out. Throughout that time, I wrote articles on this blog site which attempted to explain the technical aspects of All-Flash as the industry went from niche to mainstream. Like many technical bloggers, I found this writing process enjoyable and fulfilling, because it helped me put some order to my own thoughts on the subject. But back in 2017, something changed and my blogs became less and less frequent… eventually leading here. I’ll explain why in a minute, but first we need to talk about the title of this post.

Hardware Is Meh

Back in my Oracle days, I worked with a product called Exadata – a converged database appliance which Oracle marketed as “hardware and software engineered to work together”. For a time, Oracle’s “Engineered Systems” were the future of the company and, therefore, the epicentre of their marketing campaigns. Today? It’s all about the Oracle Cloud. And this is actually a perfect representation of the I.T. industry as a whole… because, here in 2019, nobody wants to talk about hardware anymore. Whether it’s hyper-converged systems, All-Flash storage, “Engineered” database appliances or basic server and networking infrastructure, hardware is just not cool anymore.

For a long time, companies have purchased hardware systems as a capital expense, the cost then being written off over a number of years, at which point the dreaded hardware refresh is required. Choosing the correct specifications of for hardware (capacity, performance, number of ports etc) has always been extremely challenging because business is unpredictable: buy too small and you will need to upgrade at some point down the line, which could be expensive; buy too big and you are overpaying for resources you may never use. And also, if you are a small company or a startup, those capital expenses can be very hard to fund while you wait for revenue to build.

Today, nobody needs to do this anymore. The cloud – and in particular the public cloud – allows companies to consume exactly what they need, just when they need it – and fund it as an operating expense, with complete flexibility. One of the great joys of the public cloud is that hardware has been commoditised and abstracted to such a degree that you just don’t need to care about it anymore. Serverless, you might say… (IT has aways been fond of a ridiculous buzz word)

The Vendor View: AWS Is The New Enemy

For infrastructure vendors, the industry has reached a new tipping point. A few years ago, if you worked in sales for a storage startup (like me), you found business by targeting EMC customers who were unhappy with the prices they were paying / service they were getting / quality of steaks being bought for them by their EMC rep. Ditto, to a lesser extent, with HP and IBM, but EMC was the big gorilla of the marketplace. Today, everybody in storage has a new number #1 enemy: Amazon Web Services, with Microsoft Azure and Google Cloud Platform making up the top three. But make no mistake, AWS is eating everybody’s lunch – and the biggest challenge for the rest is that in many customer’s eyes, the public cloud is Amazon Web Services. (EMC, meanwhile, doesn’t even exist anymore but is instead a part of Dell… that would have been impossible to imagine five years ago).

Cloud ≠ Public Cloud

However, nobody (sane) is predicting that 100% of workloads will end up in the public cloud (and let’s be honest now, when we say “public cloud” we basically mean AWS, Azure and GCP). For some companies – where I.T. is not their core business – it makes perfect sense to do everything in the cloud. But for others, various reasons relating to control, risk, performance, security and regulation will mean that at least some data remains on premises, in private or hybrid clouds. You can argue among yourselves about how much.

So, for those people who still require their own infrastructure, what now? Once you’ve seen how easy it is to use the public cloud, sampled all the rich functionality of AWS and fallen into the trap of having staff paying for AWS instances on their credit cards (so-called “Shadow IT“), how do you go back to the old days of five-year up-front capital investments into large boxes of tin which sit in the corner of your data centre and remain stubbornly inflexible?

Consumption-Based Infrastructure

Ok let’s get to the conclusion. A couple of years ago, Kaminario (my employer) decided to exit the hardware business and become a software company. Like most (almost all) All-Flash storage vendors, Kaminario uses commodity whitebox components (basically, Intel x86 servers and enterprise-class SSDs) for the hardware chassis and then runs their own software on top to turn them into high-performance, highly-resilient and feature-rich storage platforms. Everybody does it: DellEMC’s XtremIO, Pure Storage, Kaminario, HP Nimble, NetApp… all of the differentiation in the AFA business is in software. So why purchase hardware components, manufacture and integrate them, keep them in inventory and then pass on all that extra cost to customers when your core business is actually software?

Kaminario decided to take a new route by disaggregating the hardware from the software and then handing over the hardware part to someone who already sells millions of hardware units all around the globe. Now, when you buy a Kaminario storage array, you get exactly the same physical appliance, but you (or your reseller) actually buy the hardware from Tech Data at commodity component cost. You then buy a consumption-based license to use the software from Kaminario based on the number of terabytes of data stored. This can be on a monthly Pay As You Go model or via a pre-paid subscription for a number of years. In a real sense, it is the cloud consumption model for people who require on-prem infrastructure.

There are all sorts of benefits to this (most customers never fill their storage arrays above 80% capacity, so why always pay for 100%?), but I’m not going to delve into them here because this is not a sales pitch, it’s an explanation for what I did next.

What I Did Next

Seeing as Kaminario decided to make a momentous shift, I thought it was a good time to make one of my own. So, two years ago, I took the decision to leave the world of technical presales and become a software sales executive. As in, a quota-carrying, non-technical, commercial sales guy with targets to hit and commission to earn. Presales people also earn commission, but are far more protected from the “lumpy” highs and lows that come with complex and lengthy high-value sales cycles (what sales people call “big ticket sales”). In commercial sales, the highs are higher and the lows are lower – and the risks are definitely riskier. Since my new role coincided with the company going through an entire change of business model, the risk was pretty hard to quantify, but I’m pleased to say that 2018 was the company’s best ever year, not just globally but also in the territory that I now manage (the United Kingdom).

More importantly for me, I’m now two years into this new journey and I have zero regrets about the decision to leave my technical past behind. I’ve learnt more than ever before (often the hard way) and I’ve experienced all the highs and lows one might expect, but I still get the same excitement from this role that I used to get in the early days of my technical career.

So, the time is right to hang up the technical jersey and bid flashdba farewell. It’s been fun and I want to say thank you to everybody who read, commented, agreed or disagreed with my content. There are almost 200 posts and pages on this site which I will leave here in the hope that they remain useful to others – and as a sort of virtual monument to my former career.

In the meantime, I’ve got to go now, because there are meetings to be had, customers to be entertained, dinners to be expensed and (hopefully) deals to be closed. Farewell, my friends, stay in touch… and remember, if you need to buy something… call me, yeah?

— flashdba —

[September 2020 Spoiler Alert: I couldn’t stay away]

Flash Debrief: The End (part 1)

Seven years ago this month, I created a blog and online presence called flashdba to mark the start of my journey away from Oracle databases (and DBAing) into the newly-born All-Flash Storage industry. Six years ago this month, I posted the first in what transpired to be a very long blog series attempting to explain the concepts of All Flash to those few who were interested. At least, I always assumed it would be a few, but now here we are in 2019 and the flashdba.com blog has been read over a million times, referenced in all sorts of surprising places and alluded to by Chris Mellor at The Register. One of my articles even (allegedly) got a mention by Mark Hurd during an Oracle forecast call!

But now, for various reasons that I will explain later, it’s time to draw it to a conclusion.

Review

Wow! What a ride it’s been, huh? Seven years ago, I joined a company called Violin Memory who were at the forefront of the infant (or should that be infantile?) flash industry. At one point, Violin had a global partnership with HP to make an “Exadata-killer” machine and had a valuation estimated to be around $2bn. EMC even wrote a secret briefing document in which they said, “Violin … is XtremIO’s #1 competitor in the all-flash storage market”. Meanwhile, numerous other small flash companies were being acquired for ridiculous, crazy and obscene money despite often being “pre-product” or pre-GA.So it took a particularly special effort for Violin Memory to take that head start and end up in Chapter 11 bankruptcy in December 2016. (The company is reborn as Violin Systems now, of course – and I still have friends there, so out of respect for them I have to keep my Violin stories under wraps. Which is a shame, because boy do I have some great stories…)

Meanwhile, back in 2015, I’d decided to leave Violin Memory and join another All Flash pioneer, Kaminario – where I remain today. It’s fair to say that Violin Memory didn’t appreciate that decision, with the result that I had to spend a lot of time dealing with their lawyers. You feel very small when you are a sole person engaged in a legal dispute with a corporation who can afford an expensive legal team – you become enormously aware of the difference in spending power (although, in hindsight, perhaps Violin could have used those legal fees elsewhere to better effect). So, when the CEO of Kaminario interrupted his family holiday to call me and assure me that they would stand by me throughout the dispute, it left me with a real glimpse into the different in culture between my former employer and my current one. Also, Kaminario’s lawyers were a lot better!

The Flash Storage Wars (available now as a boxset)

The road from 2012 to 2019 is littered with the bloody carcasses of failed flash companies. From the disasters (Violin Memory, Skyera, FusionIO, Tintri, Whiptail, DSSD) to the acquisitions (Texas Memory Systems, XtremIO, Virident, SolidFire, Nimble) – not all of which could be considered successful – to the home-grown products which never really delivered (I’m looking at you, Oracle FS1). One company, Pure Storage, managed to beat the odds, ride out some stormy times and go from startup to fully-established player, although following their IPO the stock market has never really given them a lot of love. Meanwhile, EMC – the ultimate big dog of storage – was acquired by Dell, while HP split into two companies and NetApp continued to be linked with an acquisition by Lenovo or Cisco. Someday, somebody is going to turn the whole story into a boxset and sell it to Netflix for millions. Game of Thrones eat your heart out.

Yet there can be no doubt that All Flash itself has succeeded in its penetration of the previously disk-dominated enterprise storage market, with IDC regularly reporting huge year-on-year growth figures (e.g. 39.3% between Q3-2017 and Q4-2018). I vividly remember, back in 2012, having to explain to every prospective customer what flash was and why it was important. Today, every prospect has already decided they want All Flash. In fact, AFAs have become so mainstream now that, starting this year, Gartner will be merging its Solid State Array Magic Quadrant with the more traditional MQ for General-Purpose Disk Arrays. It just doesn’t make sense to have two separate models now.

So Who Won?

Good question. Was it DellEMC, the biggest company in storage and the current #1 in market share? Was it Pure Storage, who led Gartner’s most recent Solid State Array Magic Quadrant (but have it all to lose when the SSA MQ merges with the general-purpose MQ)? Or was it any number of investors and venture capitalists who managed to make money on the back of such market disruption? It’s a subjective question so you can choose your own answer. But for me, it’s very clear that there was only one winner… and back in 2012 we had no idea (although my old boss called it over a decade ago… I should have paid more attention). The ultimate winner of this war – and many other wars besides – is the cloud.

In part two – the final ever blog in this series (and possibly at all – spoiler alert), I’ll explain why I think the cloud is the ultimate winner… and why I’m calling time on flashdba after all these years. Wipe away those tears, my friends – not long now.

See also: this article apparently inspired the highly respected storage-industry journalist Chris Mellor to write A Potted History of All-Flash Arrays over at Blocks and Files. Thanks Chris, I’m honoured!