The Uncomfortable Truth about SAN/NAS
We just finished up our Storage Field Day presentation last week and one thing that was clear - there are a lot of folks in the analyst community that continue to be invested in the success of SAN/NAS.
On the one hand we understand.
There are far more companies in the SAN/NAS universe than there are in the object storage space. When you think about object storage there is MinIO, the three public cloud providers and some appliance vendors. The appliance vendors don’t really have a path forward in the cloud operating model so it is really just the public cloud players and us.
From the perspective of some of the Storage Field Day participants, there is the perception that there is a lot going on in SAN/NAS because that is who they spend the majority of their time with.
We obviously live in a different world. All we do is object storage and we mostly deal with enterprises who have already educated themselves on the technology and on MinIO.
We see a different class of problem based on who we interact with. MinIO is 100% inbound. We don’t have commissioned salespeople. Our discussions on the commercial side are with enterprises who manage many PBs of data and, due to the inbound nature of our model, generally have already sold themselves on MinIO before we start talking. That is our world.
We are not focused on small customers from a business perspective. Granted there are thousands of MinIO users at <100TB but they are not in focus for us from a commercial perspective. It is why there is a minimum of 100TB for a commercial license. The community benefits from the work we are doing at scale and with small objects but there are no features which are targeted at small workloads (unless you are aggregating data at the edge into large workloads where we are doing some interesting work).
Therein lies another disconnect.
For some of the SFD23 participants they claim phenomena that we are seeing - that of object storage eating the world, isn’t what they are seeing. Their reasoning is that most of their clients are smaller in nature - so they don’t encounter the scalability issues associated with SAN/NAS/POSIX. Some of them never will. It still doesn’t change the fact that SAN/NAS/POSIX isn’t cloud native - but let us come back to that.
Cloud Native is a Binary Choice
Are there legitimate use cases for SAN/NAS? Absolutely, and will be going forward for many years to come. They just won’t be cloud native use cases.
Do we think that fast object storage, coupled with NVMe and 100 GbE networks mean that object storage will pick up more and more workloads? Yes. And others do too. Even one of the critics of our argument agree:
While Chris Evans has artfully worded this, “In the future, if the majority of data doesn’t need the features of file storage, then object storage is the logical choice as the protocol is arguably much more efficient than NAS.”
But it’s not about the “features of file storage” is it? Truthfully, we don’t know what those features would be. Strict consistency? Present in object storage. IOPS? Present in object storage.
We think the better question is really “in the future do you want to be cloud-native or not?” If you do, you are moving away from SAN/NAS. The scale doesn’t really matter. That is a binary choice. If you want to be cloud native you are doing that with object storage.
The Uncomfortable Truth
Object storage is primary storage in the cloud. It has been since day one. Redshift is built on object storage. BigQuery is built on object storage. Snowflake is built on object storage.
The attributes of object storage make it ideal for the cloud operating model. That model has long since jumped the confines of the big three and exists everywhere - other public clouds, private clouds, on-prem and even at the edge.
As a result, high-performance, cloud-native object storage has made that leap too. It is everywhere, running every type of workload - notably the workloads that were previously the domain of SAN/NAS. Things like databases.
This trend isn’t going to stop, for the simple reason that the cloud-native movement is far stronger than the POSIX movement. The cloud operating model wins. Every analyst knows this in their heart. It is all over but the crying.
So why are we arguing over it? Is the timing inconvenient? We don’t control that. We are the messenger. The outcome was sealed a two plus years ago when it was clear Kubernetes had won. This isn’t even about “embracing change” , this is about accepting reality and moving forward.
Again, if your customer is small and doesn’t have much data, SAN/NAS will be fine for a few more years. But if you are building a modern data lake - well, that’s an object storage task - as is every other cloud-native workload.