Skip to end of metadata
Go to start of metadata
Purpose of this document
This page describes the deliverables from the Common Execution Infrastructure subsystem during Release 1 Iteration 1, structured by component.

The Common Execution Infrastructure subsystem provides an execution environment for system and user processes that enables reliable execution and scales to demand.

The Common Execution Infrastructure subsystem contains three components.

Execution Engine Catalog & Repository

This component maintains references to registered execution sites and virtual compute node configuration packages. An important responsibility of this component is to provide the tools necessary for bringing those configurations 'to life' as instantiated compute nodes, also known as 'contextualization' (see below).

Status

  • Integrate Chef (system mgmt framework) with the Nimbus Context agent (a program that runs at boot time on VMs) for improved cluster contextualizing. 
  • The integrating of Chef with the Nimbus context-agent is now working (details on this can be found at [1]).  The code now correctly goes from launching instances (via the Nimbus "cluster defining XML file") to executing "chef-solo" with the correct role-specific data (that originates in the cluster defining XML file]).
  • The Nimbus context agent was refactored into smaller modules
  • Work was done to deliver unique data roles through another generic mechanism (JSON blob in Nimbus cluster defining XML file <data> element) so the same feature was delivered without needing to modify the Nimbus Context Broker
  • A "Contextulization Scenarios" document ([2]) was created in order to provide examples of situations we might see in the working system, the main purpose being to facilitate requirements discussions. For a definition of "contextualization" see the top of that document.
  • A RESTful protocol was proposed ([3]) for the context agent and broker interactions, such an implementation would make both broker and agent implementations easier to change/maintain.

[1] http://www.oceanobservatories.org/spaces/display/CIDev/Nimbus+context+agent+-+Chef+integration

[2] http://www.oceanobservatories.org/spaces/display/CIDev/Contextualization+Scenarios

[3] http://www.oceanobservatories.org/spaces/display/CIDev/RESTful+interface+to+Context+Broker

Next actions

  1. Improve Chef+Nimbus ctx Agent to be more user friendly: Include more examples and documentation.

Jira Issues

(login required to see this)

jiraissues: Unable to determine if sort should be enabled.


Elastic Computing Services

This component provides scheduling, provisioning, and monitoring services to maintain a balanced deployment of virtual compute nodes.

Status

  • Adjusted tools to support libvirt so that more hypervisors and hypervisor features can be integrated more easily. The bridge to the libvirt library is in place but more work is needed to provide a KVM capability on par with the Nimbus Xen capability (in particular networking/security parity).
  • Created an initial administrator dashboard for Nimbus. The initial functionality allows an administrator to more easily bootstrap users with credentials to access the Nimbus service. Another feature developed is the ability for this to run "out of the box" with very limited dependencies required (Python, Java, and the answer to two questions) in order to have a running, SSL enabled dashboard. That capability will be even more important when this framework is leveraged for the installation procedure.
  • Documents describing cloud semantics and leases were produced in order to facilitate requirements discussions. See [1] and [2].

[1] http://www.oceanobservatories.org/spaces/display/CIDev/IaaS+SLAs

[2] http://www.oceanobservatories.org/spaces/display/CIDev/Job+vs.+Cloud+patterns

Next actions

  1. Work on KVM integration (especially networking/security) now that the libvirt bridge is in place
  2. Expand administrator dashboard to include installation procedures.

Jira Issues

(login required to see this) 

jiraissues: Unable to determine if sort should be enabled.

Resource Management Services

This component establishes global and standard models for the operational management (monitor & control) of stateful and taskable resources. It is the point of integration with the COI.

Status

  • Created an EC2 "query API" protocol implementation for Nimbus which enables many popular clients that only use this protocol (and its accompanying security scheme which is a unique scheme that Amazon created). See [1] for details.
  • Outlined ideas for the "ever present unit" which is the main point of COI integration, it is a system capable of making sure any requested service/module interaction messages are "fulfilled." See [2] for details.

[1] http://www.oceanobservatories.org/spaces/display/CIDev/EC2+Query+API

[2] http://www.oceanobservatories.org/spaces/display/CIDev/EPU+and+Module+interactions

Next actions

Jira Issues

(login required to see this) 

jiraissues: Unable to determine if sort should be enabled.
Enter labels to add to this page:
Please wait 
Looking for a label? Just start typing.