SwiftStack have announced a collaboration with Google Cloud Platform to take its object storage system into the multi-cloud world.
Storage vendors tend to take one of two approaches to cloud. One is to run their software on cloud infrastructure, which is easy enough because pretty much all software these days targets the x86 chipset, which is what the major clouds use in their vast server farms. This approach means the software and its functionality looks exactly the same regardless of where it runs, be it on-site, in AWS, Azure, GCP, wherever. Think NetApp ONTAP Cloud, Zadara Storage Cloud, or Microsoft StorSimple style systems, commonly called Virtual Storage Arrays, or VSAs.
The second is to connect to native cloud services and move data between them. This is the approach that SwiftStack has taken, using the de facto standard status of the S3 protocol to provide the common access method, but leaving the functional details to the specific cloud provider. This approach exposes whatever differentiation is possible in any given cloud provider’s implementation, and also frees SwiftStack from having to implement all the cloud features itself. This approach is used a lot by systems that offload data to the cloud, like Rubrik, Cohesity, Veeam, and a host of other backup style systems.
Moving the data around is still a major challenge. SwiftStack have taken a replication approach, that it calls Cloud Sync, which is the simplest way to do things. There are multi-site access systems like Panzura and Nasuni that have complex locking mechanisms, but they target a very different use-case to what SwiftStack is trying to do.
“It’s about making the data accessible independent of its access method,” said Mario Blandini, VP of Marketing at SwiftStack. “It’s why we have a single namespace, and use an open format: so you can do new things with your data without having to wait for us.” Blandini stressed the way SwiftStack uses the native S3 format of different cloud providers so that customers can choose which features of which cloud they want to use, when they want to use it.
SwiftStack customers tend to process data in pipelines. They’re not accessing the same data in multiple places at once and performing simultaneous manipulations on it; that’s an incredibly complex thing to do. Instead, they might upload a large dataset into one cloud for transcoding, before the results are pushed out to a distribution platform. Or a dataset might be uploaded to Google’s cloud for complex analysis (which Blandini says is one of the ways Google is differentiating itself from other clouds) and the metadata is then replicated back to be handed off to the next stage in the pipeline.
It’s all about using the right tool for the job, rather than ‘making do’ with whatever one specific cloud provider offers. Going all in with a single provider can provide certain benefits (a single relationship and bill to manage, for example) but there are downsides as well. SwiftStack has chosen to leave that decision to the customer.