This article is more than 1 year old

The data centre is becoming the cloud file doorway centre ...

... says cloud gateway supplier Nasuni. Quelle surprise

The advantages/disadvantages of the minivan

El Reg: How do you view the rise of hyper-converged systems, from Nutanix to EVO: RAIL, which do away with physical SANs/filer?

Andres Rodriguez: They’re good as long as you can fit the problem in the device. If what you want is pure IOPS as opposed to relatively well-managed IOPS, you’ll go to flash storage. If you want something in the middle, like SharePoint, where you have a lot of data, hyper-converged offers a better self-managed system.

If you need capacity, go to the cloud. Hyper-converged companies are like the minivan. If you can fit everything you need in a minivan, great. But if your task really needs a truck or a sports car, the minivan isn’t going to cut it.

Nutanix and the cloud are riding the same trend: virtualisation. It changed the world of block storage, networking and compute forever. In essence, hyper-converged systems take advantage of this new ability to virtualise the complete stack so they can integrate all these previously discrete functions together and then keep each other in balance.

If you need more networking, you’ve got it. More storage, it’s allocated. The cloud uses the same principals.

The problem is that this approach only works to the extent that you can represent the data to fit into a VM, and you can’t fit a bunch of files on a hyper-converged machine. As I said above, innovations around blocks are all about performance, which is expensive. The problem with files is one of capacity, a problem that isn’t solved by buying expensive flash storage. This is why the world has been separated between performance (block) and capacity (file) problems.

El Reg: Is object storage in your plans? Would you need a separate gateway for that?

Andres Rodriguez: Object storage is a new way of storing files and it’s the bedrock of cloud storage, but ultimately, what defines an object is the class of system you use to store. The object store is massively scalable and very stable, but you can’t modify the data written to it, so it’s useless for data centre file storage.

Nasuni transforms the cloud’s object store into a file store by taking advantage of snapshots, so we get the best of both worlds: high performance at the edge, and the scale and stability of an object store in the core. Plus you can access it from anywhere in the world.

El Reg: How about Ceph and Hadoop. Is data stored in these systems moving to the cloud? How does Nasuni deal with that?

Andres Rodriguez: Nasuni already has the capability to bring up instances of the Nasuni filer natively in AWS and Azure, so you can do data analytics at scale, which was not possible before. The interface is a Hadoop interface.

We have customers who are already doing this. What’s cool about this is that in the cloud you have access to compute resources that aren’t easy to get in a data centre.

For example, with search, if you’ve deployed a global file system, a customer could use the cloud to actually search though their files. It’s ideal for that, because, when you do indexing in the cloud, you’re not interfering with the performance of how your end-users are working with files.

El Reg: If compute moves to the cloud, does Nasuni have a role?

Andres Rodriguez: Absolutely. Nasuni’s mission is to give customers a place to store and access files at any scale from any number of locations across device times.

So wherever their users are, very large organisations can keep all their files with us. We provide scale and stability, including the cloud. As workloads move to the cloud, we’re there. They can access it with the same ease.

Nasuni creates the file system and then all these questions become, where do you want to access the file system from? The cloud? A mobile device? China? It doesn’t matter to us. ®

More about

TIP US OFF

Send us news


Other stories you might like