It’s not a question of if, anymore – using multiple cloud services is almost the norm for many companies. According to the 2016 RightScale State of the Cloud Report, the average cloud user is running an average of six clouds. That includes both public and private clouds, as well as clouds being used for experimentation. While speedy deployments are a factor in choosing multiple clouds and cloud providers, experts caution that it requires an application-driven strategy to succeed.

Forrester data shows that the bulk of companies are running two public clouds, but selecting the right cloud for the right app is becoming increasingly popular, according to Dave Bartoletti, principal analyst at Forrester. For example, a company might build a new mobile app on Amazon, migrate an older Windows app to Azure for modernization, and have the analytics team use Google Machine Learning. “Multi-cloud is not the future – it’s already here for nearly half of all companies using public cloud,” he said.

While getting up and running with apps is a benefit, using multiple clouds can lead to data fragmentation, security issues, migration challenges, and overall governance, Bartoletti said.

SEE: Cloud diversity: How 10 companies use the cloud 10 different ways

Choose clouds according to application

One of the ways that companies can manage multiple clouds is before the cloud is selected, by choosing additional services wisely. Bartoletti recommends following the apps to the cloud, not the other way around. “Let your app needs drive your choice of clouds, not today’s compute or storage prices – they will keep dropping,” he said. Higher value services like database, analytics, mobile platforms, and integration tools should guide the choice, he added.

Don’t let fear of lock-in guide choice

Additionally, some companies try not to use proprietary cloud services like AWS Lambda or Google BigQuery because they’re concerned about being locked into that choice, according to Mike Kavis, vice president and principal architect at Cloud Technology Partners. However, that means passing up the opportunity to use APIs that can increase agility and alleviate some of the challenges with managing cloud services, he said.

Address challenges with architecture, services at the edge

Although cloud services do offer benefits, one of the challenges lies in services like identity access management. According to Kavis, not a lot of cloud-agnostic versions exist, and many companies are forced to do redundant work on each endpoint to address that issue. However, by using a loosely coupled architecture where specific cloud provider APIs are abstracted from the rest of the code base can alleviate this problem, he said.

“For example, if the application needs to write to cloud storage, that call should be a separate function that can take a parameter that tells it which cloud provider is being requested. The function can then make the appropriate call that is specific to the cloud provider,” Kavis said.

Another challenge that crops up is cloud latency, across enterprise headquarters and the WAN to branch offices and remote sites, according to Brad Casemore, research director of data center networks for IDC. Cloud latency is something that providers are tackling, but services at the cloud edge, like cross-connection services, can help overcome this challenge in the meantime, he added.

For enterprises managing multiple clouds, going into different services with an application-driven strategy can assist with management challenges. Companies should also look to APIs and edge services to address challenges. As more applications move to the cloud, it’s only a question of when before all enterprises are juggling a combination of cloud services.

Also see:
Google’s Hölzle on why customers shouldn’t have to care about cloud infrastructure
Don’t get bullied into public cloud: Private cloud is growing, says report
Gut check on Oracle’s journey to the cloud
Switching clouds: What Spotify learned when it swapped AWS for Google’s cloud