Back in 2014, the world was a different place. It was the height of the big data era, and the cloud was still hotly debated. Not many people bet that the cloud would come to dominate compute workloads, let alone become the de facto data store.
MinIO founders, however, were not most people -- so they took that bet. Coming from a background of distributed file systems, storage, and open source, Anand Babu Periasamy, Garima Kapoor and Harshavardhana invested in creating open source cloud storage software to address the demands of unstructured data growth in the cloud.
Today, MinIO Inc., creators of the MinIO multi-cloud object storage suite, announced that it has raised $103 million in Series B funding at a $1 billion valuation. The investment was led by Intel Capital with participation from new investor SoftBank Vision Fund 2, and existing investors Dell Capital, General Catalyst, and Nexus Venture Partners.
We caught up with Periasamy to discuss MinIO's journey.
The persistent layer of the data stack
Periasamy previously founded a startup called Gluster, which built an open source distributed file system, and was acquired by Red Hat. He has been in open source for as long as he can remember, and around 2014, it became clear to him and his MinIO co-founders that there was a need to be addressed in data:
"Anything you do around data is here to stay. We picked the persistent layer of the data stack that is storing and managing all the data. And the opportunity for us was the public cloud."
MinIO founders believed the storage for the cloud is going to be the storage for everybody, and that storage is object storage. Of course, each cloud has its own object storage. So MinIO had to take their bet further, although that was an easier choice to make.
What they did was they built their software to be 100% compatible with the leading cloud's object storage -- Amazon's S3. But MinIO does not just run on Amazon's cloud. It runs on Microsoft Azure, Google Cloud, Azure, VMWare Tanzu, OpenShift, SUSE Rancher, bare metal, the edge, hybrid cloud, multi-cloud -- pretty much anywhere.
MinIO's bet was that AWS will be very successful in convincing the industry of the right way to build infrastructure, but it's highly unlikely that AWS will be the only infrastructure around. And in that scenario, having too many standards or APIs was not going to work. It took discipline to stick to that decision, not add support for this or that API, or legacy file systems, and just stick to S3:
"A Grand Unified Theory of Storage sounds nice on paper, but these are completely independent problems. The whole idea about object storage was to let go of legacy stuff. If I bring back the legacy stuff, then it's not a plus, it's a minus. I was just disciplined to say "no" a lot. And that's what allowed us to get here", Periasamy said.
Of course, building a product around someone else's API can be challenging. Jonathan Symonds, MinIO CMO, said that gratitude is owed to Amazon for the S3 API because they changed the developer mindset in terms of how storage is done. He also emphasized the power that comes from MinIO's open source model.
That has enabled MinIO to harden their implementation of the S3 API "in a way that no proprietary software vendor could ever do", Symonds said. Periasamy also noted that there is 100% feature parity between the community and the enterprise edition of MinIO. MinIO has a dual licensing model, whereby its Community edition is licensed using GNU AGPL v3, while the Standard and Enterprise editions leverage a commercial license.
Periasamy believes this is the clearest possible model. AGPL is not commercially licensable, but that actually helps MinIO make money by granting a commercial license for the Standard and Enterprise editions. At the same time, MinIO does not have to backtrack from a permissive license such as MIT or Apache to an open core model. This seems to work for MinIO, to the point of convincing investors: we don't often see AGPL-based unicorns.
Amazon S3 as the one object storage interface to rule them all
Still, there is the elephant in the room. Is not MinIO worried about having to build on Amazon's S3 API specification? What if Amazon decided to go Oracle on them and tried to prohibit them from using the S3 API? There seems to be a precedent on that now, but Periasamy did not refer to that in addressing the question.
That would be just as much, or perhaps an even bigger issue for Amazon, Periasamy said. S3 is an API, not a standard, and Periasamy commented that the API definitions are more like a religious text: There are many ways to interpret them, and Amazon's own SDKs interpret them differently between different languages, or even within different versions of AWS' Java SDK.
But this is not just about the legal rights to use, he went on to add. Amazon could simply keep changing S3, even if they never sued anybody using the AWS S3 API. If they did that, they could upset the industry, but also hurt their own customers. Periasamy believes the S3 API is pretty much converged, and it's not going to change.
Additions, however, are fair game. If Amazon keeps adding, MinIO will keep adding too, Periasamy said. He strongly believes, however, that S3 already has the right functionality. Periasamy went on to add that Amazon has been kind to MinIO and to everybody else. He believes the reason is that they have a similar mindset.
If S3 API becomes popular, it means it will result in more applications being Amazon ready. And Amazon's belief, as per Periasamy, is that they can build the best product compared to anybody else. If that's the case, and S3 API indeed becomes a standard, the industry will gravitate towards AWS and they will win anyway.
MinIO has the same core belief too. The difference is, Periasamy elaborated, Amazon thinks that they will be the world's data center. MinIO thinks that they will build better software, and turn all infrastructure into an object storage service.
Still, if ever there were any issues with Amazon, MinIO would, and could, proceed to publish their own standard, with the weight of its massive community behind it, Periasamy said. But it's actually in the best interest of all parties to cooperate.
Periasamy also claimed that the industry is moving away from legacy formats towards object storage en masse. Case in point -- HDFS, Hadoop's file system. He confirmed that MinIO is displacing HDFS and Hadoop on a massive scale. He noted that Hadoop's architecture is not in line with cloud requirements; however, he gave credit to Hadoop for having kickstarted the big data era.
What is in line with cloud requirements is Kubernetes, and MinIO is playing along. Kubernetes was not there when MinIO started. MinIO itself is stand-alone and can run from Raspberry Pi to all kinds of high-end servers. But as Kubernetes grew, it became clear to MinIO that "they got the ideas right" and they took a serious bet on Kubernetes.
With or without Kubernetes, however, MinIO has some impressive metrics to show: Annual recurring revenue grew by over 201% in 2021, while customer count grew by over 208%. MinIO also boasts 31,000 GitHub stars and 16,000+ MinIO Slack community members.
As for future plans, Periasamy outlined MinIO's focus as focused on educating the market at an even broader level. He believes that object storage has the potential to see even greater growth, and promises MinIO will make it their business that this happens.