Tintri storage strategy shoots straight at public cloud

Tintri storage is taking a cloud approach with its VM-aware arrays. Co-founder and CTO Kieran Harty said Tintri is adapting an API-based web services approach to help IT organizations deal with "pressure from the CEO to do things differently" by applying the economics and scale of the cloud.

Harty was Tintri's original CEO until Ken Klein took over in 2013. Prior to launching Tintri, Harty spent seven years as an executive vice president of engineering at VMware, leading the delivery of ESX Server, VirtualCenter and VMware desktop virtualization products.

Tintri VMstore hybrid and all-flash arrays are based on a deep integration with VMware. The Tintri arrays operate at the virtual machine (VM) and disk level, replacing conventional file abstractions with VM-aware storage. Harty gave SearchCloudStorage an update on Tintri's emerging cloud strategy and planned product rollouts in 2017.

You say Tintri's cloud strategy is based on automated web services. What does it entail and how does it change the way people use Tintri storage arrays?

Kieran Harty: A lot of our customers have a huge impetus to do cloud initiatives. The public cloud model has taken off in a big way with our customer base. They want their existing infrastructure and applications to be available with the agility of the public cloud. They don't want to rewrite their applications or redo their compliance for a cloud environment.

Are you trying to avoid getting pigeonholed as a storage-only vendor?

Harty: We're still an all-flash array vendor. Obviously, our VMstore flash storage itself is invaluable. There's a lot of complexity associated with implementing good storage in terms of performance, cost and reliability. But storage is table stakes.

Kieran Harty

We're taking a web services approach that allows you to automate everything. The basis of web services is that everything is defined using APIs [to provide] the right level of abstraction.

How would a customer use Tintri storage to build an on-premises cloud?

Harty: Today, more than one-third of Tintri customers are using our platform to build their own cloud with varying degree of cloud capabilities. They use our web services architecture to automate common tasks like provisioning new virtual machines or applying Quality of Service policies, to scale-out while optimizing the location of every individual virtual machine and to apply predictive analytics to anticipate their future need for capacity and performance. 

When we asked our customer advisory board a year ago whether they need connections from their own data center to public cloud (S3, Azure, etc), they said no. But there has been a transition within the last year where they decided they do need it. So I think it's going to be a gradual process.

More companies want a cloud implementation on premises behind their firewall. Does Tintri plan to emulate the Amazon cloud computing model?

We're trying to provide you [with] the agility of the public cloud, but within your own data center.

Harty: Yes, we're taking a very similar approach to the enterprise environment. Compute has been a web service since VMware introduced its hypervisor products. You also have Microsoft Hyper-V and containers emerging. Network is becoming a service with things like VMware NSX and Cisco's ACI (Application Centric Infrastructure). But storage has not [developed as a service] at the same pace.

Most storage vendors aren't using a web services approach. They design their architectures using the same physical concepts you have within the traditional data center, where you create LUNs and volumes and have people provision and interpret the stats associated with a particular VM. The model we have is one in which all your workflows are fully automated, including the storage, compute and networking that is provided by other vendors. We're trying to provide you [with] the agility of the public cloud, but within your own data center.

How quickly will you roll out support for public clouds?

Harty: We plan to start integration with AWS later this year, which will give you the ability to send and retrieve granular VM snapshots with Amazon. Gartner is using the term 'cloud-inspired infrastructure' to describe this uptake and it's very consistent with what we see as well. We're starting with AWS because it's the big gorilla. We'll add Azure support if we see a customer demand for it.

What concerns do you have about the recent AWS outage? What redundancies does your cloud strategy include to ensure customers have access to their Tintri storage?

Harty: The AWS outage validates our belief that customers should have a multifaceted cloud strategy. We allow customers to have data in AWS and [also] have data on premises via Tintri enterprise cloud. In the case of an outage, Tintri storage would still be able to offer data on premises via snapshots. Tintri has the ability to replicate data to multiple different sites. If you care about the availability of your data, you can replicate it to multiple sites as well as to AWS, which will give you high availability even during an outage.

What new Tintri storage products are you developing as part of the virtualization strategy?

Harty: The basis for what we do is everything being done in automated fashion. You can create a snapshot on a VM that talks to us, the storage vendor. You can set [Quality of Service] on the VM which talks to us, the storage vendor. And you can integrate those workflows in an automated, self-service way. Very distinct in terms of the capabilities we provide. And it's based on the concept that everything we do is based on a web service.

We've introduced support for VMware vRealize Orchestrator that allows you to orchestrate more general storage workflows. We've had asynchronous replication, but we just introduced synchronous replication capabilities. That allows you to have a disaster recovery capability where all the data is sent to a remote site. You can have two VMstore arrays up to 60 miles apart. Data gets written to both arrays. If there's a failure of the primary array, the secondary array takes over.