Archive:Tools/VirtAPIDevelopment

From FedoraProject

(Difference between revisions)
Jump to: navigation, search
(remove from virt category, it's out of date)
m (bad hr!)
 
(2 intermediate revisions by one user not shown)
Line 18: Line 18:
  
 
There is a need for a very low overhead API for colllecting, distributing and processing resource utilization data. eg, Ganglia but with a formal API accessible to applications. A prototype of such a system is in development under the name [http://spectre.et.redhat.com Spectre] .
 
There is a need for a very low overhead API for colllecting, distributing and processing resource utilization data. eg, Ganglia but with a formal API accessible to applications. A prototype of such a system is in development under the name [http://spectre.et.redhat.com Spectre] .
 +
 +
[[Category:Virtualization archive]]

Latest revision as of 15:32, 19 January 2009

Stop (medium size).png
This page is out of date
Please see libvirt for up to date information.

This page deals with development of virtualization APIs for Fedora. It is principally interested in libvirt at this time.

[edit] libvirt todo

All development on libvirt is independant of the Fedora Project. This todo list simply indicates the priority (wish) list for libvirt functionality needed by Fedora

  • Hot-plug/unplug of devices. In particular to allow CDROM devices to be switched while a guest is running
  • Support for managing inactive Xen domains. eg by scanning for config files /etc/xen
  • Remote management. Allow use of libvirt APIs from a remote machine - currently limited to on-box management
  • QEMU backend. Allow management of QEMU instances using same APIs as for Xen.
  • KVM backend. Could likely re-use 95% of the functionality from QMEU backend.

Anyone interested in volunteering to work on these items should make themselves known to the libvirt mailing list.

[edit] Resource monitoring

There is a need for a very low overhead API for colllecting, distributing and processing resource utilization data. eg, Ganglia but with a formal API accessible to applications. A prototype of such a system is in development under the name Spectre .