Difference between revisions of "Jacket"

Revision as of 07:09, 14 February 2016

Overview

More and more enterprise are thinking about how to leverage the usage of private cloud and public cloud. Public cloud have very good scalability and have cost advantage in many cases, while Private cloud have more security assurance and have more feasibility for IT administrator. It is a hard decision to choose just one of them. A traditional Cloud Broker can provider a basic multi-cloud management function, but it is not a good experience for the administrator, since they still feels like they are using multiple separated clouds with a unified UI style. What's worse is that they need to learn one more type of operational mode and API besides the existing clouds APIs.

A hybrid cloud solution is required, which will provide smooth inter-cloud management which makes you feels just like managing one cloud, and it should provide the existing standard API and operational mode which makes you shorten learning curve and easy to integrate with industry existing services.

The target of Jacket project will solve the problem with OpenStack. It will be an adaptor for the different clouds, and it will be called by the services in OpenStack, for example nova, cinder and so on. In this way, OpenStack will manage the different clouds and offer the standard OpenStack API just like one cloud. At the same time, users can enjoy a series of the 3rd part tools which are based on OpenStack.

Use case

Cross cloud deployment

A typical web application is composed with 3 different tiers, the load balancer tier, the web front tier, and the app backend tier. Users have a private cloud based on OpenStack or VMware and so on, which just has less resources but more secure than public clouds. So they can use the standard OpenStack API to launch the web service's backend virtual machines in private cloud, launch the load balancer and web front virtual machines in the public clouds.

Cloud burst

In some cases, user would not like to put anything in public cloud if not necessary. User would like to put all parts of the application in private cloud usually, and they would like to scale the web front end in case of there is a 'Cloud Burst' comes. For example, I have a online shopping mall in my private cloud, the system resource and internet bandwidth is enough in normal case. But there will have a bottleneck in big festival period, so I would like to scale out my application to public cloud in that period, and I would like to scale in my application back to private cloud after the big festival.

Challenges

How to shield the functional differences between clouds

There are huge differences between the API models of the different provider clouds. For example, VMware vCloud Director can't support to create the bootable volume through the image, and it can't support the snapshot/backup of the volumes. But OpenStack and AWS can support these festures. So it is the first challenge to offer the unified standard OpenStack API model to manage the different provider clouds.

How to unify images and flavors' management

The provider clouds have the seprate image storehouses, and they have the different demands for the images. At the same time, there will be the same images, which are the same OS version in the markets of the provider clouds, and the particular images which just can be used in the .

The provider clouds has defined the different virtual machine flavors for the users, which are differ in thousands of ways.

How to unify images and flavors' management just like one cloud?

How to import the existing resources in the provider clouds

Users maybe have a lot of resources in the provider clouds before they use this solution. How to import the existing resources will be a challenge.

Architecture

Architecture layer 0

The advantages of this architecture:

• One OpenStack can manage the OpenStack private cloud and other heterogeneous clouds at the same time.
• Only one jacket-driver in nova/cinder/... services will be offered to manage the different clouds with different configuations in jacket-dirver.
• The adaptor for the different provider clouds can share the CPU/memory resouces.
• Jacket service can be deployed as an auto-scaling group, so that it can scale out easily.

To do list