EMC’s ViPR software-defined storage platform will not be dependent on other suppliers writing adapters to link their arrays to it.
If they don’t write them, EMC will, said EMC president of advanced software Amitabh Srivastava (pictured), speaking to Computer Weekly about EMC’s vision for ViPR as an operating system-style environment to which heterogenous storage could connect.
He also gave hints about ViPR’s likely charging structure and EMC’s view of storage in the software-defined datacentre.
ViPR will allow customers to build large pools of web-scale storage from heterogenous storage arrays, EMC or third-party storage supplier arrays, or even commodity hardware. ViPR uses its own Object Data Services that can be accessed via Amazon S3 or Hadoop distributed file system (HDFS) application programming interfaces (APIs) to enable analytics of data where it resides on the storage media under its management.
While EMC outlined the main contours of ViPR at its Las Vegas event, some questions remain about cooperation with other storage array makers and how EMC will charge for ViPR. Computer Weekly caught up with Srivastava to ask those questions.
ComputerWeekly.com: The success of ViPR depends, to a large extent, on being able to connect to other suppliers’ arrays, and that means having adapters? Which other suppliers are going to write adapters? Which of them has so far? Which of them have you had talks with?
Srivastava: We have designed ViPR to handle heterogenous storage – EMC, non-EMC, commodity – and we already have adapters for NetApp and will have them for IBM and Hitachi Data Systems (HDS) arrays.
I was at Microsoft and worked on Windows Server products and [Microsoft’s cloud platform] Azure, and that’s the model we have in mind for ViPR, with connections from any device via a device driver that anyone can write – EMC, other vendors, third parties.
It doesn’t matter if the storage vendor doesn’t write adapters, because anyone can write them. If customers need them, they will get written.
It’s really simple to write them – a case of translating between ViPR and the array. We wrote the NetApp adapters in two days. We expect IBM and HDS adapters to be next, as that’s what customers are demanding, and we will write those too if we need to.
ComputerWeekly.com: How will you charge for ViPR?
Srivastava: We haven’t announced what charging model we will use yet, although the biggest feedback we’re getting from customers is that they want it to be simple, such as by capacity and pay-as-you-go. We will announce it very soon, and it will be aimed at enterprise and service provider customers.
ComputerWeekly.com: How does ViPR fit into the software-defined datacentre model?
Srivastava: We really believe the enterprise wants to go in the direction of the software-defined datacentre, which I think comprises four key elements:
- Software designed for massive scale, cloud operations and openness.
- The ability to abstract heterogenous storage arrays and so preserve the value of existing investments.
- Policy-based management at the abstraction layer.
- Building cloud data services into the software for all underlying arrays.
The cloud requires a flexible datacentre, and the compute and networking elements are there already. The only piece that’s missing now is software-defined storage.
We are building that software-defined storage – taking heterogenous storage, block, file, object and commodity storage, and putting an abstraction layer on top of it.