Comment Last week, your correspondent attended SFD10 and I met with a lot of interesting storage startups. Most of them have a good level of integration with VMware (who doesn’t today?), but one in particular built its entire strategy around it: Tintri.
Tintri and its VM-aware storage
The company started in 2008 with the idea of providing a storage system particularly optimized for virtualized environments.
At the beginning it was VMware-only, but now it also supports Microsoft Hyper-V and KVM (RedHat and OpenStack) with a similar set of functionalities.
As far as I know, Tintri has been the first to provide VM granularity for almost any feature available on its storage system; Snapshots, replica, QoS and so on. Forgive me for the oversimplification but it was like having VVOLs before VVOLs were even conceived.
Furthermore, a few weeks ago, the company announced new “scale-out” functionalities, much improved analytics and self-optimisation capabilities to get the most out of your VMware infrastructure.
I wrote “scale-out” in quotes because it’s not a real scale-out storage system, as it is for other vendors like Solidfire or Isilon. This is more like a federation of different storage arrays managed as a single system, and with the addition of smart data and workload placement features to help keep it balanced and efficient – resembling VMware DRS on the compute side. Don’t get me wrong, the product is very well designed and offers some very neat features, it’s just that I don’t think about it as a scale-out solution.
In any case, marketing and details apart, Tintri has done an amazing job and I really like it (you can watch a demo here). It eases up the sysadmin workload and simplifies his/her life without particular limits or constraints.
I’m not saying it’s perfect, though. A 1.0 version is still just 1.0, and further automation and optimisations will see the light with subsequent releases. To be honest, this is also to be expected; no one would trust the system enough to give it complete control over data and workloads before spending some time to understand how it works.
I like Tintri a lot – but it's not all perfect
Like I said, Tintri has a fantastic product and it can be considered a great solution when it comes to virtualised infrastructures – but I think that’s also its biggest limitation.
The VM-aware functionalities available on a Tintri system are now table stakes for many other players. We could argue that Tintri does it better, but most of the vendors are catching up very quickly. Also, the analytics part is no longer news and many vendors offer similar features, or even more if we think about Nimble Storage InfoSight, for example. And what about scale-out? Nearly all HCI vendors offer it. On the other hand, Tintri doesn’t support non-virtualised workloads which can be a limit for external storage.
In the last four to five years we’ve seen several new startups doing very well in the hyper-converged infrastructure space. They offer a seamless user experience and similar granularity to what Tintri promises. And, if you look carefully, Tintri offers most of the HCI benefits through a separate box but, to the end user, the idea of further simplifying the infrastructure by having fewer boxes is even more compelling.
At the end of the day Tintri has two types of competitors: HCI vendors on one side, who offer a similar experience but with less complexity, and external storage vendors on the other, who offer more flexibility in terms of OS and workload support. All of them are now capable of offering VM-level granularity for the vast majority of the features and/or cloud-based analytics.
The VM-aware storage (sales) dilemma
VM-aware storage also has another problem: the buyer. Who is Tintri’s first buyer? The storage or virtualisation team? Storage people don’t like limited scope storage systems. They want flexibility, parity of features between different OSes and a system that can provision storage resources when and where needed. Virtualisation people adore VM-aware storage but HCI goes a step further for them: why buy an external storage system if they can have everything they need directly from their servers?
Customer size is another issue. If a small customer buys an external system today they buy only one and they want it unified (multi-protocol and multi-purpose) or they go for HCI 100 per cent. Single purpose storage systems are for larger organisations. Yet, in this case, we fall back to what was discussed above.
Closing the circle
I’m really keen on Tintri and their storage system and, as far as I know, so are their customers. But VM-aware is now one of the many features required in order to be competitive. Tintri has the best in class implementation regarding VM-aware functionalities, but it’s not enough for today’s needs.
If the goal is simplification, hyper-convergence is perceived as a further simplification when compared to external storage. In the other cases it lacks the features to be compared with other external storage systems. I think this says a lot about VM-aware storage, doesn’t it?
I want to close with a small comparison here. Last week I also met Nimble Storage, but the same argument can be applied to Tegile and others. Nimble has a wonderful analytics tool (probably still the best on the market), transparent migration functionalities similar to what is possible with Tintri, scale-out and a not-so-distant VMware granularity for their features.
I’m using Nimble as an example because they started in the same year, 2008. Today Nimble has more than 8,000 customers and, if I understood correctly during the SFD session, Tintri has only 1,000. I have to say that Tintri is pushing much harder on sales now than in the past, but it’s also true that if the market thinks Nimble struggles at times, then what should we think about Tintri? ®