An interview with Coldago research analyst Philippe Nicolas has revealed what he sees as vendor and product selection choices that in his view weaken Gartner’s Distributed File Systems and Object Storage MQ.
Read the Q&A below and see what you think.
Blocks and Files: Should distributed file systems and object storage be viewed as a single category?
Philippe Nicolas: Hmm, it’s a good question. What is true is both address unstructured data but many applications can use one and not the other, even if the access method is standardised. At the same time, we see more and more vendors offering both interfaces. Clearly it creates a challenge when you need to analyse the segment.
If we consider these two access models as one category, Gartner has to select products that do both to avoid a penalty for only file or object vendors. But why should a vendor be penalised when it delivers only one interface, especially when that can be a very good one?
Considering the two as one category invites us to make the same point we have made for years: Gartner considers one product for some vendors and multiple products for others, and therefore creates an unfair or unbalanced comparison. So the real question is, do we compare one product or do we compare vendors?
Some suppliers, such as Pure Storage and Scality, are combining file and object storage. Shouldn’t analysts do the same? And if not, why not??
And you can add Caringo (now DataCore), Qumulo, DDN, Dell, NetApp, VAST Data or Cloudian to extend the list; I probably even forget a few. This is a general answer that demonstrates once again that differentiators across offerings are reduced year by year. It’s also a sign of maturity. Having check boxes ticked in RFPs does the job but product behaviour is very different.
How vendors implement their access layers really differs. But it also confirms the merger between offerings — because it’s essentially two access methods to access the same unstructured content.
Also, you can merge the category but what about pure object storage or pure file storage products/vendors? Does it mean we need separate sub-MQ for each category with the presence of players who deliver the individual access layers? I think this is where other analysts reports come into the game, and users must consider several of them to form their own ideas and opinions.
Purists would tell you that object storage is more than just an interface and they’re right, but nobody cares today about internal design, especially when products expose both interfaces. Many users ask their vendors: “Could you expose my content on a file server via S3?” and the reverse as well.
But all these products are far from equal when we look at access methods. Do you really compare native NFS access built on object layers and vice-versa? Of course, it can provide some flexibility but users’ experience shows very diverse capabilities and realities.
And lastly, the problem with grouping the two is that some pure file or object players are sanctioned. And this is a paradox — you can be a very good product in one category but badly positioned in the global quadrant. On the other side, having the two, let’s say with average capabilities, provides some artificial advantages.
With flash hardware and better-designed object software accelerating object storage to filer-level performance and so to satisfying primary storage roles, aren’t the two access protocols (file and object) merging?
Flash is used in object storage for metadata for a long time and it was too expensive for data for large configurations. But the reality was also that some object storage products didn’t receive any perormance gain by using flash for data and several of them had to adapt, change and update their software to maximise the gain. And then flash pricing went down so it created some extra opportunities.
Your point is interesting. I remember a recent study by one vendor claiming that object storage with flash can do primary storage. In fact, primary storage is only determined by its role and not by a technology. Many people limit primary storage to block storage and it’s a very narrow view of the sector. Primary storage is where data is generated and thus it’s active and considered hot data. It supports production and sustains the business. With that in mind we understand that it can be block, file or object, whether HDD, flash, SCM or full DRAM lies underneath.
On the other side, secondary storage is a protection level, needed to protect the business and support IT in its mission. Data is not generated there — it’s copied from the primary level. This secondary level is full of inactive data — cold and even fixed or reference data. Here we see also some block, file or object access systems.
Your question remark confirms, once again, that object storage has become an interface in people’s mind.
What is your view of the general relevance and usefulness of Gartner’s Magic Quadrant for Distributed File Systems and Object Storage?
I like it, I like that exercise, it’s good that such tools exist with several other ones to invite users to read and analyse several of them, understand context and criteria to form their own opinion. We just regret that some visible players are not listed and that Gartner didn’t accept or consider points many other people do year after year.
Even if we understand the criteria chosen by Gartner, it is always a surprise to not see some players as they refuse to be listed or because Gartner eliminates them. Look at the trajectory of VAST Data in the market — not having it listed is pretty bizarre and makes this report a bit incomplete.
What about open-source? What about MinIO, clearly the number one object storage by the number of instances running on the planet?
And the reverse is also true in this MQ. I’m pretty sure that all readers were surprised to see some brands on it this year.
How should and could IT buyers find out MQ-type information about the distributed file systems and object storage suppliers if the Gartner MQ is rejected?
Hmm, there is no one source of information and I invite buyers to make their own search of similar reports and analysis to build their own matrix with their own criteria as a mix or union of these documents. Honestly they already do this for RFPs; it’s just an extension. When they need to research the state-of-the-art in a domain, they have to do it. A good source is a few key information sites like yours, StorageNewsletter, TechTarget, Speicherguide.de and a few others that go beyond posting press release and analyse things. And lastly, if buyers can speak directly with users who have already deployed and adopted solutions, they’ll get excellent inputs.