Kieran Harty, Tintri.
Why did you start Tintri?
I’d been a VMware for seven years and I had some ideas for starting a company. I settled on storage because I saw a big opportunity. I connected with Mark [Gritter, co-founder and architect] through a mutual friend.
We’re quite unusual for a start-up in that we’re still solving the problem that we set out to solve. Namely, that customers get a lot of benefits from virtualisation on the compute side – typically reduction of physical servers by a factor of ten or more. But storage is the dirty little secret around virtual storage, because it puts far more stress on storage.
Why is storage a problem in virtual environments?
Customers have had to pay just as much or even more for storage in a virtual infrastructure. But as well as the cost there are also often performance problems, and it’s a much more complicated environment. That’s one of the major things about this market, it would take any company two to three years to get a product to market. You need to build a file system that is very reliable, and incredibly performant. You need a team of at least 20 people just to start.
Content from our partners
You’ve taken on venture capital funding…
We’ve raised a lot of VC. Because development takes a while you really don’t have any choice. We took some on in 2008 and since then we have done a further two rounds. We’ve raised $60m in the last five years, and today we have 180 customers and 110 staff.
Why tackle the storage market? Storage incumbents like EMC, IBM and HP all have massive R&D budgets and huge advertising and marketing budgets too?
That’s true, but the thing that has encouraged me is that I think we’re seeing the biggest change in the storage market for more than 40 years. Specifically, advances around flash and 10 Gigabit Ethernet. The technologies that EMC, NetApp and so on have were typically designed 20 years ago. The existing vendors are not advantaged by the pace of change. And to adapt they have to cannibalise their own business. It would take EMC four or five years to innovate to the point we got to in two or three. There’s a good chance for smaller companies to be successful.
What about the technology – it’s flash-based I understand?
There’s actually a mixture of SSD flash and SATA. But the innovation is really around the VM [virtual machine] awareness. Traditional storage was all built around physical concepts, but in the virtual world these become limiting. It’s very difficult to get good performance and to move things around. Our storage is VM aware. It’s a bit like comparing early Windows with users’ knowledge of things like registries and file systems, to an iPad where such things are completely hidden. We have a similar level of abstraction. You really don’t need to know an awful lot about storage to configure our products.
Doesn’t flash have a much shorter mean time between failure than more traditional disk?
Some of our intellectual property is around exactly that, the wear-out rates around flash. The reason flash wears out is because every time you do a write to it, it writes a lot more data than you think. Eight kilobytes can turn into several megabytes due to this issue of write amplification. We’ve gotten around that, so that if you write eight kilobytes it’s only eight kilobytes. That gives the drive a lifetime of over five years which is long enough.
You’re not the only ones doing flash: I’ve talked to companies like Whiptail, Coraid…
Our primary differentiator is that we’re not taking a storage model from 20 years ago and re-architecting it for flash. We built it from the ground up to work in a virtualised environment. Today 65% of all workloads are virtualised and in the cloud space it’s almost all. Across the board it will be 80- to 90% in the next few years.
You are certified to work with the VMware hypervisor. Any plans to support any others, like Microsoft HyperV or Xen or anything?
We are working on Microsoft HyperV and probably over time we’ll support others, possibly KVM.
What is your headline claim, like, ‘twice as must storage for half the price’?
We had one customer using traditional storage and they had a quote for $2.5m which was negotiated down to $1.25m. We sold them storage for $350,000 which is twice the speed and takes up one seventh of the rack space. It was installed in less than an hour when the former storage was going to take two days at least.
What next for Tintri from a design perspective?
The goal is to operate on larger and larger estates of VMs. To be the storage used in infrastructures of hundreds of thousands or even millions of VMs over multiple sites. To manage VMs in Tokyo, New York, London and still manage them in a very simple way. That’s always been the vision and we’re on the way.
