About those Cloud APIs...
About those Cloud APIs...
Join the DZone community and get the full member experience.Join For Free
Learn how to migrate and modernize stateless applications and run them in a Kubernetes cluster.
Originally authored by Sebastien Goasguen
There has been a lot of discussions lately within the OpenStack community on the need for an AWS API interface to OpenStack nova. I followed the discussion from far via a few tweets, but I am of the opinion that any IaaS solution does need to expose an AWS interface. AWS is the leader in Cloud and has been since 2006 -yes that's seven years- Users are accustomed to it and the AWS API is the de-factor standard.
When Eucalyptus started, it's main goal was to become an AWS clone and in 2012 signed an agreement with Amazon to offer seamless AWS support in Eucalyptus. Opennebula has almost always offered an AWS bridge and CloudStack has too, even though in total disclosure the interface was broken in the Apache CloudStack 4.1 release. Thankfully the AWS interface is now fixed in 4.1.2 release and will also be in the upcoming 4.2 release. To avoid breaking this interface we are developing a jenkins pipeline which will test it using the Eucalyptus testing suite.
Opennebula recently ran a survey to determine where to best put its efforts in API development. The results where clear with 47% of respondents asking for better AWS compatibility. There are of course developing official standards from standard organizations, most notably OCCI from OGF and CIMI from DMTF. The opennebula survey seems to indicate a stronger demand for OCCI than CIMI, but IMHO this is due to historical reasons: Opennebula early efforts in being the first OCCI implementation and Opennebula user base especially within projects like HelixNebula.
CIMI was promising and probably still is but it will most likely face an up-hill battle since RedHat announced it's scaling back on supporting Apache DeltaCloud. I recently heard about a new CIMI implementation project for Stratuslab from some of my friends at Sixsq, it is interesting and fun because written in Clojure and I hope to see it used with Clostack to provide a CIMI interface to CloudStack. We may be couple weeks out :)
While AWS is the de-facto standard, I want to make sure that CloudStack offers choices for its users. If someone wants to use OCCI and CIMI or AWS or the native CloudStack API they should be able to. I will be at the CloudPlugfest Interoperability week in Madrid Sept 18-20 and I hope to demonstrate a brand new OCCI interface to CloudStack using rOCCI and CloudStack ruby gem. A CloudStack contributor from Taiwan has been working on it.
The main issue with all these "standard" interfaces is that they will never give you the complete API of a given IaaS implementation. They by nature provide the lowest common denominator. That roughly means that the user facing APIs could be standardized but the administrator API will always remain hidden and non-standard. In CloudStack for instance, there are over 300 API calls. While we can expose a compatible interface, this will always only cover a subset of the overall API. It also brings the question of all the other AWS services: EMR, Beanstalk, CloudFormation... Standardizing on those will be extremely difficult if not impossible.
So yes, we should expose an AWS compatible interface, but we should also have OCCI, CIMI and of course our native API. Making those bridges is not hard, what's hard is the implementation behind it.
All of this would leave us with Google Compute Engine (GCE), and I should be able to bring back some good news by the end of september, stay tuned !!!
Published at DZone with permission of Mark Hinkle , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.