Cloud computing offers scalability, operation cost and flexibility to the companies.
All of these seem like good reasons to believe that moving into the cloud will have immediate benefit, but each one throws up its own challenge for companies, says Martin Percival (pictured), senior solutions architect, Red Hat.
Scalability without control tends to lead to wasteful overuse of resources, projects grow in scope very quickly and are left consuming resources, long after testing is complete. Cloud ‘creep’ needs measurement to bring it back under control, adds Percival
“That same measurement very quickly shows that without planning, public cloud usage can be more expensive than originally believed," he added.
"Not only that, but comparisons of potential costs between public cloud vendors is a difficult task and any idea of finding some kind of ‘cloud resource brokerage’ is impossible to achieve when the best prices come from longer contracts,” said Percival.
Flexibility is non-trivial thing to deliver, he added. Applications vary wildly in their needs.
What may run fine in the controlled environment of a more monolithic style data centre, could raise different configuration issues for every public cloud vendor, effectively preventing easy, flexible migration.
Not only that, but green field applications written specifically for the public cloud, can quickly become as locked in to the APIs of an individual cloud vendor as they would have been in the days of proprietary software stacks.
“We are at a point where a combination of technologies can start to solve some of the bigger questions posed by cloud infrastructure. First, cloud management. For this to be successful it needs to be able to encompass both the public cloud vendors and also any in-house private clouds, whether built from raw VM infrastructure or using something more sophisticated like OpenStack,” said Percival.
To deliver wide value, it needs to be able to be implemented without invasive changes to existing infrastructure and should be able to interrogate and control any existing management layers where desired (and permitted). While the term “single pane of glass” is much overused, it should provide visibility of all elements of a company’s cloud and expose the full deployment of any apps across these elements.
He adds that some form of automation tool can help to ensure that applications are deployed in a repeatable manner, no matter where in the cloud they physically run. Like the management tool, it should understand the needs of all the different public clouds and should be extensible to handle new situations as they arise. It should work alongside the management tool as the ‘delivery arm’ of the combined toolset.
“We should look to build software architectures that transcend the API’s of any one vendor,” he said. The rise of containers as an app development and delivery mechanism is a promising direction.
In particular, the use of Kubernetes as the orchestration engine for those containers is a very successful cloud vendor-neutral way to ensure the flexibility that businesses need.
“With these pieces in place, we can finally start to realise the broader promises of a truly hybrid cloud environment, that we can both measure and control to meet the needs of any kind of business application,” Percival added.
