buchachon - Fotolia
- Dave Raffo, Editorial News Director
You don't hear much talk of the cloud anymore.
Sure, everyone is talking about clouds. But we no longer consider cloud one large, homogenous entity under the banner of the cloud. We've come to understand that all clouds are different, whether they're hosted by the likes of Amazon, Microsoft or Google or on premises by IT staffs. Each cloud has its strengths and weaknesses.
Now the talk is of multiple clouds, or multi-cloud for short. IT no longer investigates how it can use the cloud, but how it can best take advantage of the various cloud options available to hone a multi-cloud strategy. (The exception is Amazon, of course, which wants the world to think AWS is the only cloud that matters.)
A multi-cloud focus is a good thing, because it makes people realistic about the value of clouds. But like all new technology buzzwords, it can be confusing. When storage vendors talk multi-cloud strategy -- and storage vendors are all talking a lot about multi-cloud these days -- what exactly do they mean?
One thing to know is multi-cloud doesn't only mean you can move your data from one cloud to another to cut costs. In the early days of clouds, there was a lot of focus on if one cloud vendor raised or dropped its prices, would it be possible to move data to a cheaper alternative? That's a moot point now, because prices across the major public clouds aren't much different for similar services. And moving data would negate any cost savings, anyway.
Multi-cloud also doesn't mean running a single application across different clouds -- not yet, anyway, and maybe not ever.
That said, there are good reasons to use more than one public cloud. A multi-cloud strategy lets you take advantage of a service that one cloud provider excels in for one use case and a different cloud for something another provider does best. You may want to move data to use Amazon Redshift or TensorFlow inside Google, for example. Developers may need to access resources stored in different clouds as if they were in different on-premises locations. And, of course, it helps with data protection to not have all your data in the same place.
The trick is to take advantage of these services without creating silos that require different sets of management tools.
I've spent a lot of time talking with storage vendors about how their customers are using clouds recently.
"Vendor lock-in is something enterprises are conscious of," said Patrick Harr, CEO of cloud NAS pioneer Panzura. "If it's all in one cloud, you have a 'Hotel California' problem. But also, a customer may want to take advantage of Google's machine learning service; perhaps it's better-suited for them than something in Azure. So we'll move the data to take advantage of that service and then bring that result back to the customer."
Storage vendors are giving their applications features that facilitate multi-cloud storage. Next-generation distributed file systems support a single namespace to access data in any cloud. Open source support helps eliminate cloud lock-in. Common management and policies must also fit across different clouds.
"Cloud is like an operating system, just like Mac and Windows," said Joe Arnold, president and chief product officer of cloud object storage vendor SwiftStack. "And each cloud has its own services and functions that make it unique. Amazon is the gorilla, but we're seeing that Azure and Google have unique functionality. How do I leverage those features, even if I have some data on premises and some in Amazon? If data is only accessible for one location, you can't take advantage of those other services in another cloud."
Arnold said a single namespace needs to look at data across clouds as if it was "a bunch of objects in a single bucket of data. You don't need to know some objects live in Google, some live in Amazon, some live on prem."
Data protection and a multi-cloud strategy
Regardless of how many or which clouds you use in your multi-cloud strategy, one thing remains the same: It's all about protecting and managing data wherever it resides.
"Customers care about the storage, but they care more about the data. How can they better utilize that data? How can they better repurpose data to generate new revenue?" said Eran Tamir, vice president of product for cloud object storage vendor NooBaa. "Say a customer has data in AWS and wants to write new data to Azure because of a new service he wants to consume. That customer may need mirroring and not just a link between clouds. There has to be flexibility in how we manage and move the data."
That common management must include on-premises data, too, because most data still resides there. Enterprises often like to manage on-premises data like it's in a public cloud and share the load across private and public clouds.
Anand Babu Periasamy, CEO of open source object storage software vendor Minio, said while developers are embracing public clouds and use Minio for that, most of its revenue comes from on-premises implementations.
"The private cloud is where we're seeing dollars," he said. "Public cloud is great for adoption, and you cannot ignore the development phase. They often go to public cloud to do their development. Public cloud gives us a great way to get adoption, but we get money from the private cloud. Enterprises are saying, 'We want our storage to look like public cloud.'"
- How should our team manage data stored in the cloud? –SearchCloudComputing.com
- Big Data and Analytics Applied to Oil and Gas –ComputerWeekly.com
- Making multi-cloud work –ComputerWeekly.com
- Computer Weekly - 5 November 2019: The benefits of API-first software ... –ComputerWeekly.com