Infrastructure Software is Dead

It took me too long to realize it, but now I am certain. Infrastructure software is dead. And with it, numbered are the days of any company whose core business is pinned to selling licenses or subscriptions to infrastructure software bits. It feels as though the entire industry is stuck in a self-perpetuating cycle of trying to justify its own existence. And if we keep it up, we’ll be dead too.  

Let me take a step back and set the context. We came late to the OpenStack game. At that time software startups like Piston and Cloudscaling, as well as established companies like Red Hat and HP were well underway in their race for the “king of OpenStack software” title. Fast forward five years and we are behind the largest OpenStack deployments on the planet, able to win deals from a super competent rival with 20 years of open source history.

Now I’d love to tell you that it’s all because Mirantis OpenStack software is so much better than everybody else’s OpenStack software, but I’d be lying. Everybody’s OpenStack software is equally bad. It’s also as bad as all the other infrastructure software out there – software-defined networking, software-defined storage, cloud management platforms, platforms-as-service, container orchestrators, you name it. It’s all full of bugs, hard to upgrade and a nightmare to operate. It’s all bad.

But none of this matters, because today customers don’t care about software. Customers care about outcomes. And the reason Mirantis has been successful is because, despite ourselves, outcomes are what we’ve been able to deliver to our customers by complementing crappy OpenStack software with hordes of talented infrastructure hackers that made up for the gaps. We didn’t win because of the software. We won because we’ve been shouldering the pain associated with turning OpenStack software into customer outcomes.

Seventeen years ago, Salesforce.com transformed the business application space by pioneering the SaaS model. Ten years ago AWS did the same for infrastructure. Neither of these transformations were about the software; they were about the software delivery model.

Infrastructure is deadWhen Salesforce.com launched, Siebel was a 2 billion dollar software company. At the time, Siebel could not have beaten SFDC by writing a better version of its CRM software, because SFDC did not innovate on the software; it innovated on the delivery model that abstracted customers away from the pain of operating (forever crappy) enterprise CRM software.

In hindsight it is obvious, right? Why is it then that after AWS disrupted the infrastructure delivery model a decade ago, today we still see infrastructure startups and veterans trying to capture market share by shipping a better version of infrastructure software? Would it not be the same as Siebel trying to beat SFDC by building “better” CRM, while continuing to ship it on CDs?

Analysts constantly ask me: “Is OpenStack mature for the enterprise?” Just a couple more years and it will be mature, right? Wrong. OpenStack is quintessential enterprise software. No enterprise software delivered as packaged software bits will ever be mature for today’s enterprises that are hooked on the cloud delivery model. Especially not the cloud software itself.

At Mirantis we entered the market with a services-first approach, focused on helping customers build and operate OpenStack before ever releasing an OpenStack software distribution. To this day, many still don’t understand the business model. “Do you want to be a product company or a services company?” – we get asked. “Is AWS a product company or a services company?” – I reply. Cloud is about redefining the notion of “product.” Software is the product of yesterday. The product of today is a combination of software and, more importantly, a service to operate that software, getting the customer to their desired outcome.  So when Mirantis is discounted as “primarily a services company“, I reply: “You bet we are!”

18 responses to “Infrastructure Software is Dead

  1. Your title and your argument, while provocative, are incorrect.

    While it is true that SaaS took market share away from on premise business applications it did not eliminate them. In fact, the very article you link to make your point makes the opposite point, that Salesforce did not disrupt Siebel but rather that Siebel failed on its own due to a more nuanced set of factors (“The mythical story of “the David” Salesforce slaying “the Goliath” Siebel that has been propagated over the years may play well inside the halls of Salesforce or in various industry reports, but the truth is far more pedestrian. Both companies were headed toward an eventual collision. But Siebel succumbed well before such an epic battle occurred, to a much bigger problem — a global recession. And once Siebel was acquired, few of the people who made the company successful stayed at Oracle.”).
    Anyone who has thought deeply about the question of infrastructure software has an equally nuanced view. The world will continue to consist of a mix of private and public cloud deployment, a mix of open source and commercial software, and a mix of products and services. Anyone who claims that only one model will win the day and wipe out everything else likely has a vested interest in and a myopic view of the model that they are promoting.

    1. Thank you for the comment. In general, I agree that the world is never black or white and yes we’ll see mix models long into the future. Mainframe business is still very lucrative for IBM.

      So yes, SFDC did not eliminate Siebel immediately. It did disrupt the market though by introducing a low touch delivery model for enterprise software. Similarly, in infra stpace cloud did not eliminate infrastructure today, but I am talking about the trend and near term future. On Siebel vs. SFDC topic specifically I think this graph says it all

  2. Nice article but, if Mirantis wouldn’t have gotten tons of $$$ from Intel, now it would be in a very bad situation. Being a services company required a much bigger magnitude than Mirantis’… Thus, let’s wait until the Intel $$$ are over!

  3. Aren’t you are making an observation about delivery of OpenStack infrastructure, and then over-generalizing that to a sensationalized headline about infrastructure software in general? To determine whether this is indeed a leap in your logic, I’d pose three questions to you:

    1. From where / whom are your customers (or indeed anyone using clouds) obtaining the operating systems / distributions on which they run their workload instances in the cloud?
    2. From where / whom do they get security updates, bugfixes, feature enhancements etc. for those operating systems / distributions?
    3. Wouldn’t you label those operating systems / distributions and corresponding updates as “infrastructure software”?

    The same considerations apply not just to the instances inside the cloud, but to the OpenStack infrastructure itself, which still needs some kind of Linux to run on.

    You are of course right that there are major paradigm shifts of delivery models occurring within the industry, but to declare infrastructure software dead when it is one of the fundamental building blocks of the new models sounds like pretty obvious marketing spin, as the previous commenter already noted.

    Disclaimer: I work for an infrastructure software company. Our “products” have been services-on-software for almost 24 years (in fact we were arguably the first to pioneer a enterprise business model selling services on top of FL/OSS), and rather than dying off we’re currently doing better than ever.

    1. Adam – thanks for the comment. We love SUSE so don’t mean to offend, but… yes – IMO, Linux distribution business is definitely at risk in the transition that the industry is going through. Linux server has two value sides – one side is hardware facing i.e. HCL/verified drivers, the other is app facing i.e. APIs and verified user space packages bundled into a distro. No prominent public cloud vendor today uses a commercial linux distro as a host/hypervisor, they all build and maintain their own flavor – which takes care of the hardware facing side of value. When it comes to the app facing side – most Linux distros today (optimized and certified for traditional enterprise apps) actually are not suitable for running distributed, net-new apps so people with custom packages to create custom images which invalidates support – which takes care of the user facing side of Linux distro value. BTW, this is also Red Hat has two versions for RHEL – one RHEL proper (slowing moving but supported for years so Oracle can run on it etc.) and a different RHEL branch that is not available standalone but is used to be “co-engineered” into their OpenStack and RHEV builds. So yes, commercial Linux distros are definitely in a tough spot.

      1. Thanks for the reply Boris. No offence taken 😉

        On the hardware-facing side, you are almost certainly right about the prominent public cloud vendors, who are so huge that they can afford to maintain their own in-house distro just for running their cloud infrastructure. This is also true of some of the largest private cloud users. But what about all the smaller public cloud vendors, and most private cloud operators? Maintaining infrastructure software is not their primary business, so it’s more economical for them to delegate this to a vendor – good news for Mirantis OpenStack, SUSE OpenStack Cloud, and our other competitors, all of which run on top of commercial Linux distros.

        On the app-facing side, I think the crux of where we disagree was captured when you said:

        > “most Linux distros today (optimized and certified for traditional enterprise apps) actually are not suitable for running distributed, net-new apps so people with custom packages to create custom images which invalidates support”

        Firstly, we optimize and certify our distros for both traditional enterprise and new cloud-native apps. Secondly, our customers can build custom images including custom packages, and still have the image supported by us. I know you said “most” not “all”, but even if SUSE was the *only* vendor with these offerings (which I doubt we are), that would drive an increase in demand for our services rather than a move towards a future where infrastructure software is dead.

        Whether it’s hardware-facing or app-facing, you are of course right that there will always be companies who prefer to build their own rather than buy, but infrastructure software is still infrastructure software even when it’s custom-built. So I stand by my belief that your article title is conveniently misleading 🙂

        If our current performance constitutes being “in a tough spot”, then I hope we have many more “tough spots” in the future, and the OpenStack market is big enough that I’m sure you’ll enjoy some of them too 😉

      2. What is your take on coreos, which is designed to be the universal linux host delivery platform. Beyond a certain point the diversity of hardware will diminish, as per foot compute fabric goes commodity, then the need for diversifed os configurations will die away.

        What happens when coreos replaces the EFI bios side on the hardware side, and your system is delivered in situe with ready to go container host.

  4. You said:
    “Do you want to be a product company or a services company?” – we get asked. “Is AWS a product company or a services company?” – I reply. Cloud is about redefining the notion of “product.” Software is the product of yesterday. The product of today is a combination of software and, more importantly, a service to operate that software, getting the customer to their desired outcome.

    FWIW, this is exactly what we’re seeing in the market.

  5. You said:
    “Do you want to be a product company or a services company?” – we get asked. “Is AWS a product company or a services company?” – I reply. Cloud is about redefining the notion of “product.” Software is the product of yesterday. The product of today is a combination of software and, more importantly, a service to operate that software, getting the customer to their desired outcome.

    FWIW, this is exactly what we’re seeing in the market. The problem is, it’s a lot harder to scale than a simple software solution.

  6. If Cloud is the new politics, then infrastructure companies are currently being whipped into the new governance, and they are doing it with a weird unsure smile of their faces, looking backwards at their share holders and presenting some “Is this cloud?” middleware crap. In 2007 Apple made Nokia irrelevant by changing the user interface, Nokia were fermenting at the top of the cell phone industry when Apple delivered a hardware and software device – a phone, Nokia were traded as junk bonds later that year and Apple took over the music and movie delivery and rental business, retail shops closed down – because of a new user interface. Infrastructure owners – print media, moves, music etc got some new middlemen- itunes, netflix, etc. without much fanfare. OpenStack is to AWS as Android was to iPhone, but OpenStacks failure is like going back to 2007 and showing people the new iphone but them deciding to stick with their Nokia

    1. The problem of Nokia was that
      they failed to create their own software ecosystem,
      and then they joined another failed ecosystem.

      But I believe that back in 2007, most of people would have sticked with their Nokia 😉

  7. We need to make better enterprise software, that applies to Open Stack.
    If everyone needs to buy your services to setup openstack, we are doomed. We need software which works out of the box and we know what I am paying for. This is the classical problem with all open-source projects. That’s why sometimes open-source software is expensive than properitry software.

  8. Your opinion in brief is software for infra is dead, but you do the service side biz which has the potential. I think the word ‘service’ has different meaning there. Today’s service must be focused on standardization of everybody’s common needs, little on the support for somebody’s special needs.

  9. Why can you have “hordes of talented infrastructure hackers that made up for the gaps.” but you cannot have hordes of talented engineers to make the infrastructure product work ? Why can only services be delivered by cloud ? Why can’t products be delivered by cloud ? Your arguments are not convincing.

  10. I understand that blog’s titles must be controversial.

    And I agree that:

    1)
    infrastructure becomes commoditized and software defiend,
    but not dead.
    There are more opportunities then ever before (see 2 and 3).

    2)
    enterprises will have to run their workloads using cloud model sooner or later.
    The traditional way is not scalable, managable or economical for big data, Internet of Things etc.

    3)
    enterprises always wanted solutions to their problems, and not puzzles of components, layers and dependencies.
    Cloud and the underlying infrastructure is not an exception.
    It is just very complex so the problem is more visible.

    An average enterprise will not assemble PCs, compile kernel, tweak kvm, write librados or SDN apps.

    I’m gland to see that Mirantis wants to rebadge itself from OpenStack software into cloud service
    company.

  11. Oh my. Another “something is dead” article. Apple is Dead. PCs are Dead. Gaming on PCs is dead. Dead dead dead. In the end, this is all about ensuring hosting companies like AWS continue to draw subscription services no matter what. The problem is eventually people are going to get subscription burnout and get sick of having everything as an overpriced monthly fee. And when companies are repeatedly seeing that a good small datacenter with Cisco UCS and some high speed flash based hyper convergence products can do the job much cheaper, I think this current cycle of outsourcing will die like EVERY cycle of outsourcing we have experienced. Making everything as a service doesn’t fix anything, except moving all the braintrust of infrastructure to centralized companies like RackSpace and AWS. But do we really want all the IT brainpower outsourced, taking ALL the control out of our hands? This is all madness and not good planning. It’s bandaging the real problem without addressing the true beast. Bad software.

Leave a Reply

Your email address will not be published. Required fields are marked *

NEWS VIA EMAIL

Recommendations

Archive

On-Demand Webinar
Introducing Mirantis Cloud Platform