Why should we be looking to multi-cloud?
“There will not be only one cloud…but rather a handful of large cloud firms with companies taking a multi-provider approach,” AWS CEO Andrew Jassy told attendees of the Gartner Symposium last year. One year later his prediction is already coming true, with more than 80 per cent of firms now moving to multi-cloud (according to Rightscale’s 2018 State of the Cloud report).
“It’s about choice,” Scality’s Brad King said at Computing‘s recent Cloud & Infrastructure Live! event. “Being able to choose where I put data, in which country and the services that accompany it.”
The main benefit of a multi-cloud approach is the freedom to choose the most appropriate cloud on a per-application basis, whether that is public or private. However, that doesn’t mean switching based on day rates: considering how the cost of bandwidth affects data egress, that vision of the future is still some way off. Storing data across clouds, though, is a very viable application today.
Scality examined the cost of storing one petabyte of data with different cloud providers over three years. The initial benchmark was $750,000 to store that in AWS, or $1.5 million to store and replicate.
The firm found that by storing data in one cloud and replicating it in another, the price can be brought down quickly. “At most, it costs about five per cent of TCO to replicate data from one site to another – so cost is not a valid argument [against multi-cloud].”
A combination of AWS and Azure is cheaper than just AWS, while using Azure and a “cheap and cheerful” provider like Wasabi or Backblaze for backup costs almost the same as a single instance of AWS.
Analysing data in the cloud is another way to avoid paying high prices. “If you have applications that push data to a cloud then do analysis of the data, and you delete the data and just keep the results, you don’t pay the egress fees… We see a number of our customers doing these kinds of things.
“Probably one of the first ones to do this was the New York Times, which pushed [images of] all of their old newspapers up to AWS; they did character recognition on all of the data, brought down this little tiny bit of text and just threw away all of the images that they’d pushed up to the cloud.”
Multi-cloud has many other potential applications, such as scalable analytics, burst computing and content distribution. It also brings its own challenges, such as efficiency of retrieving data; data location; and storage costs.
To deal with these issues, Scality has devised four ‘pillars’ of multi-cloud data management. They are:
Have one API across all clouds;
Keep native data format when pushing data to cloud;
The ability for multi-cloud management, pushing data from one cloud to another using their resources, rather than your network;
Being able to search metadata.
The final pillar is the most important, said King, as it enables you to search for data across all of the clouds that you are using. Scality had a customer recently who knew that it would, in the future, need to split off all of its UK-based data from the rest, because of Brexit. Tagging data as it is stored enables you to make policy decisions and change it down the line.