Introduction

• 31 page slide presentation is available. Please press the download link. The presentation comes with audio and can only be utilized once you download it. An alternative link provides the presentation with audio online
• 8 page slide presentation

Cloudmesh client allows to easily manage virtual machines, containers, HPC tasks, through a convenient client and API. Hence cloudmesh is not only a multi-cloud, but a multi-hpc environment that allows also to use container technologies (under development).

Client based. Cloudmesh client as the name indicates is a client based toolkit that is installed and run on the users computers. This also includes an add on component to the cloudmesh client which is a portal. Hence we distinguish the client that contains most of the functionality, as well as a portal that can access the functionality through a locally maintaine Web portal. Important to note is that the user manages its own credentials and thus security and credential management is done directly on the users machine instead through a hosted Web portal. This increases the security as access to any credential is managed by the user and is not part of a credential management system.

Layered Architecture. Cloudmesh client has a layered architecture that allows easy development of new features. This also allows contribution by the community while developing integrated and smaller sub components. Figure A depicts the various layers. A resource abstraction layer allows the integration of a multitude of resources spanning HPC, Containers, and Cloud resources. (At this time we focus on Openstack and Slurm resources. We are working on reintegrating resources such as Azure, AWS, Maui, Moab, and others which we previously supported, as well as new resources such as docker).

Figure A: Cloudmesh layered architecture.

Management Framework. Cloudmesh client contains a management framework, and its components are depicted in Figure B. cloudmesh allows easy management of virtual machines, containers, and the data associated with them. We are currently developing a choreography framework that leverages Ansible, chef, and heat. All of the functionality is easily usable through a command shell that also can be used from the commandline, and a Python API. IN future we will be providing a REST API.

Figure B: Cloudmesh component overview.

Database Agnostic. Cloudmesh contains some state about the resource and environment that a user may want to use. The information is managed in an database abstraction that would allow storing the data in a variety of databases such as SQL and MongoDB. At this time we have chosen SQLite to be the default database as it does not require any additional setup and is universally available on all operating systems without change.

Command shell and line. Cloudmesh contains a command shell allowing scripts to be developed and run. However we designed the command shell in such a way that each command can also be called from the command line. Through the cloudmesh state machine the state between command shell, command client, and the portal is shared.

Cloudmesh Client Portal. Previously, we distributed cloudmesh with client, server, and a portal components in one package. This however turned out to be to complex to be installed for some of our less technically skilled user community. Thus we split up the install into two independent packages. The cloudmesh client and the cloudmesh portal. The portal provides some elementary features to manage virtual machines and HPC jobs. At this time the portal is considered to be alpha technology. Just as the client the portal is to be run on the local user machine in oredr to allow increased security by managing the credentials locally rather than on a server.

Cloudmesh Two Factor Authentication. We have an exploratory project in place that looks at the use of Yubikeys for cloudmesh, client and cloudmesh portal.

Cloudmesh Comet. We are actively developing the client interface for SDSC’s comet supercomputer allowing bare metal provisioning. The interface reuses cloudmesh components and technologies while interfacing with the comet cloud REST interface. The goal here is to manage virtual clusters.

Where to go next?

What to read next may depend on your interest. Certainly you want to install cloudmesh while following the

Installation information

Next we recommend that you get familiar with the concept of defaults in cloudmesh. After that you have several options:

• If you are interestted in clouds such as Openstack read the Section Cloud Commands
• If you are interested in Comet read the the comet command manual.
• If you are interested in HPC read the Section HPC Commands