Is open the new closed?


This post is nuanced and closely tailored to some of my thoughts on the increasingly popular question: “How to best monetize OSS in the enterprise?”.

After listening to the most recent ~30min long a16z podcast, I wanted to take the opportunity to document some point-in-time thoughts. Based on my observation, many view points promoted in the podcast are also starting to generally pick up across the software industry – in other words, this is less of an immediate reaction and more of an encouraging point of reference to get my thoughts out.

The podcast “Monetizing Open Source (Or, All Enterprise Software)” was published earlier this week by a16z’s always great Sonal Chokshi hosting two experienced folks in the enterprise software and OSS world(s?): Martin Casado, now a GP at a16z, formerly co-founder of Nicira which started the SDN movement and was acquired by VMW for $1.2B in 2012 just 5 years after founding and now represents in NSX a grand fifth of VMW’s total revenue… and James Watters, founding BusDev/evangelism lead on the Cloud Foundry project at VMW and now SVP Products and Engineering at Pivotal, the company behind supporting CF and commercializing it along with Spring and a variety of other OSS projects). Getting these two folks together to talk about OSS monetization strategy, approaches and trends promised to be an interesting conversation worth listening to and I was certainly not disappointed. There were a lot of really insightful points and ideas provided.

However, when I finished listening to the podcast (ok, during, as well), I got a bit riled up inside after hearing a number of comments I strongly disagree with – particularly oriented around the mechanics of monetizing OSS / how to go about doing so. Since I believe Martin and James are broadly respected in the industry, I am sensitive to all the OSS entrepreneurs/founders/aspiring creators who listen and heed. I really love helping people and sharing my own lessons learned without reservation and the motivation for writing this comes from that. Having said this, much (all?) of the podcast was painted on the backdrop of Pivotal’s proclaimed $270M+ in license (subscription?) revenue for PCF (Pivotal Cloud Foundry, their distribution of CF++). I mostly disagree with comments below by James* – my take on several quotes from the podcast (mostly verbatim and not taken out context) is offered below.

*I should say: I probably philosophically agree with ~80% of what James publicly states and happen to have spoken with him 1:1 on a few occasions, found him to be consistent in his views there and a highly passionate individual. We have reached an important point of progression in our industry where Pivotal is becoming a shining example of potentially “how to do it right” when it comes to monetizing OSS and my response is in that context.

Optional / potentially superfluous reading: For myself, I also happen to have some years of open source and enterprise software experience. Maybe not as lucrative as Martin and James, though (working on it!): in 2010 I helped Talend go to market in the US as they were expanding internationally from France where they got started. Talend IPO’d last year and is one of the major commercial OSS successes listed in my index. After a couple years of serious hyper-growth OSS learnings there, I joined TIBCO (bit of a false start, but I was super impressed by their tech, wanted to learn and dig into their stack and vision) while TIBX was in the very early days of navigating their response to OSS pressures from upstart competitors like the one I was at previously… after that, I joined Enstratius as they were getting acquired by Dell Software – Enstratius was one of the first multi-cloud management vendors (if you are not familiar with this space, see slide 9 here). I helped Enstratius expand their presence in the US and learned a ton about how cloud management tool abstractions were starting to evolve.. then Docker happened. As 2013 progressed, I became very excited about Mesos (thanks to my friend John Willis for referring me to the Mesos paper earlier in 2013). I reached out to Mesosphere to see if I could somehow help and join them. After some exciting months of consulting and supporting Mesosphere early in their commercialization journey, Kubernetes was released, a project which has really fundamentally changed the conversation and patterns / standards / trends in this ever so rapidly evolving area of modern application development and container cluster management. This brings us to the “present day” – I was very fortunate to help start the first Kubernetes company, Kismatic, and ended up selling it to Apprenda (a company on slide 8 in the PaaS category, interestingly competing with CF, but we won’t get into that here). I also founded KubeCon, the Kubernetes community conference – personally ran the the first two events in SF and London and remain close to things even after we donated it to the Linux Foundation / CNCF. I draw all my learnings about growing and understanding OSS communities from this. KubeCon expects to have 5,000+ attendees this year across events in Berlin and Austin. My reason for sharing this is that the past 6+ years of my career have been really saturated with OSS learnings and I certainly have opinions that have evolved over that period.

OK – here we go:

“Catering to OSS users is a false horizon!”

This idea was discussed quite a bit. The basic concept here is as follows… once your OSS project starts to get a lot of momentum with users (developers?) and the adoption curve continues up and to the right, as the (a?) company behind that project, you really shouldn’t focus on those users – at least in terms of monetization.

So, I find this kind of direction very dangerous. If you ignore / don’t pay enough attention to / iterate around your OSS users’ feedback early on (even/especially from the beginning), you can very easily miss out on improving the UX in certain areas that will matter greatly later on in enterprise scenarios where you have a chance to sell a product and scale a business model. One example is an OSS project optimized for extreme scale in every way from design decisions early on, but as that project starts to creep into enterprise environments that very often start with small deployments, these initial optimizations can really hurt monetization experiments.

“Up-selling OSS users to a commercial product simply doesn’t work!”

Less black and white. There is some truth to this, but taken bluntly this can really be drastically misinterpreted. Ultimately, OSS acts as a hyper-efficient distribution model for software development, go-to-market, validation, product-user-fit cycles, and more. Integrating the behavioral conditions of critical mass adoption in OSS projects that represent networks of users in enterprises from a “bottoms up” perspective very often lends well to “up-selling” to a commercial offering. Talend, for e.g. would see several users of the project in an organization, the user company would organically see the value and ultimately engage Talend directly on their own (dynamic and natural lead generation pull minus marketing/sales spend.. magic!) to investigate the “enteprise version” which wrapped the disparate workflows of the fully OSS version into a managed and central server. This model can be implemented in a wide range of ways.

“Catch a trend and drive monetization from that instead of commoditization / OSS got a false start by being focused on commoditization!”


“Relegating OSS to commoditization instead of being transformation / mega-trend oriented is short sighted!”

This plays against the “conventional” path OSS projects have historically taken when endeavoring to capture market share / value in commercial entities like Red Hat with Linux, JBOSS with their app server, etc. Typically, there would be these big scary and vulnerable incumbents (Microsoft, IBM, Oracle…) with super high prices and market dominance. Some keen and experienced entrepreneur would then come along and go “why doesn’t everyone have access to XYZ capabilities for free instead of paying seven figures for a license?”. Said founder would then go and build a business around scaling support/education/product around her OSS offerings that provided 90%+ feature parity compared to the proprietary alternative and goodies for the latest abstractions and protocol trends, minus the bloat and baggage of the incumbent. Once this project and company scaled, the entire model would be based on the idea of “commoditizing” the pricey alternative. In the podcast, this model is seriously poopoo’d on.

For sure, there are casualties of this model. Not all commercial OSS has succeeded. Neither has closed! Commoditization is a very valid way to monetize OSS and a great initial path to disintermediate incumbents – it MAY not be the best long term path to continue growing towards achieving market dominance.. no, you need continual innovation for that. And new category creation. However, this growth and aggregate mass that flows from the commoditization path affords infinitely valuable insights into innovation rates that incumbents can only dream of achieving. This is why many commercial OSS companies end up getting acquired by the incumbents they were trying to commoditize from the start. Oh well… :(

What wasn’t really discussed (which is, I think, the best way to monetize OSS while leveraging the commoditization aspects where appropriate): true category creation based on major shifts. Not just exploiting shifts and framing the offering/value around that. Actually going out and defining, building and setting the stage for a new layer in the stack that needs to exist. A crucial point of consolidation in skills, standards, etc. That is the best way to monetize OSS. See Cloudera’s “Data Hub/Lake”… See Nicira’s “SDN”… etc.

“OSS is now the default buying criteria for enterprises!”

In such context, this is not something to be said in passing without really acknowledging / respecting the past. The acceptance of OSS and behavioral patterns around “OSS is the norm for new initiatives” is a totally recent / new phenomenon! All the heavy ground work laid by commercial OSS companies of the past 5-10 years has brought us to this point. At Talend, we often had to fight for months trying to convince purchasing groups that large OSS licenses (subscription oriented and not purchased out of CAPEX budgets like perpetual / proprietary licenses of the past – in many cases still current) were viable and the way of the future for limiting risk, ensuring predictability and removing vendor lock-in from a TCO (total cost of ownership) perspective. We are now living in an era where subscription-based licenses based on OSS product are the norm. This is after a huge amount of education, legal acceptance and industry progress.

I have also written about some of these enterprise adoption dynamics in the past.

“Distributing OSS as a service is far better than just offering support / distro / packaging!”

While true, this is an opaque statement. What does it really mean? Platform9 has recently popularized this model of offering a managed control plane for highly complex and fast moving OSS bits while having the customer deploy lightweight agents on their own hardware/VMs with a safe port open to communicate with Platform9’s “OSS-aaS” layer for managed K8s and OpenStack – fully insulating customers from upgrades/patches/version conflict, etc. In my view, something like this comes closest to the idea of “OSS-aaS” without going all the way… Unless Pivotal claims to be doing something similar, which I know is not the case, then founders/creators listening to this advice will find it hard to unpack what is meant by distributing OSS in this way. Taken literally, OSS value is captured by cloud providers like AWS/GCE/Azure by offering fully hosted services based on OSS projects (see GKE, Redshift, etc.), but IMO this is a highly orthogonal model to what James is espousing.

In spirit, I do agree with this idea – details matter, though. :)

“UI as closed is crucial to monetization of open infra components – this is where to differentiate!”

Keeping a UI closed as core to how you differentiate your OSS product is probably one of the most slim / risky ways in which to lock down your value. It has to be much more multi-dimensional and faceted across a wide range of areas that matter to an enterprise customer: security, compliance, SLA management, encryption, AuthNZ, advanced operator controls and more are common areas where differentiation layers atop OSS.

“Being the sole supplier of the product/offering is key to price control!”

This comment came from the topic of doing enterprise license transactions and mitigating risk of multiple vendors bidding on a deal in a market where more than a single “enterprise” vendor exists for a given OSS project.

While it is true that being a “single vendor OSS” company optimizes your ability to capture the value from the project and exert price control, the size of your pie overall is much smaller than if you were the “dominant” vendor for a project with a real ecosystem. A potentially longer post is required to expand on this. Let’s leave that for some other day…

“Many OSS companies have died because they never got account control .. transactionality vs strategic advisory relationships!”

Maybe so, but the evolutions of OSS commercialization and ACV (average contract value) progressions over time that I’ve observed in the industry show deals starting small and growing to being strategic in size as more adoption increases.. not dissimilar to SaaS mechanics – certainly not starting out as big bang deployments of yesteryear’s proprietary and monolithic Oracle.

“If you tried to go out and raise money around an OSS middleware company in ~2008, no one would have invested – no mega-trend/architectural shift!”

Talend raised $4M A in 2006, $3.5M B in 2007 and $12M C in 2009. They IPO’d at $560M last year. Now valued at $860M+ ($100M RR). That’s a major transformation and shift, if you ask me.

MuleSoft raised $4M A in 2006, $12M B in 2007, $12M C in 2010. They will IPO at $1.4B this year. Transformational.

Pentaho raised $13M A in 2005 and $24M B in 2006. HDS acquired them for north of $1B in 2015. Transformational.

“Successful OSS strategy is Enterprise software strategy!”

This implies nothing has changed. It has. Big time.

“Community strength often doesn’t affect CIO priorities!”

Community is not only crucial, it is imperative to the maximal success of many projects that are becoming category defining. This greatly matters to CIOs. See Kubernetes. It basically came out of nowhere and surprised many. Google gave it tons of credibility, but it came way later than Mesos, CF…. after billions were poured into those projects and communities. Now, ask yourself why it is dominating the adoption space, conversation and (IMO) soon to be dominating the enterprise and API surface area of everything in its path? Community strength. Highly relevant to CIOs. I talk with and know many.

There is another dimension here…Is your community distributed and truly diverse? Is it dominated by a single vendor? These aspects also matter.


Now read this

On data center operating systems

After strangely being blocked from commenting on the InfoWorld article below, presumably by its authors, I thought I would post my thoughts here on my blog.... Continue →