VCS Instantaneous Notification and Fast Failover via IMF/AMF
VCS 5.1 SP1, released 01-Nov-2010, offers a new feature called IMF (Intelligent Monitoring Framework).
- IMF provides the following two benefits:
1) Instantaneous Notification for resource state changes (processes and mounts going offline or online), thereby strengthening the Fast-failover use-case.
2) Reduced CPU consumption by VCS agent processes when the number of resources being monitored is high, thereby providing significant
performance benefits in terms of system resource utilization.
- The IMF feature can also be used along with the CFS (Cluster Filesystem) product to provide fast failover in the range of a few seconds for mission-critical environments where every second of failover and recovery time matters.
- Without IMF, it can take upto 1 minute to detect online state changes and upto 5 minutes to detect offline state changes.
- In case you are wondering about the difference between IMF and AMF,
AMF (Asynchronous Monitoring Framework) is the kernel module that provides the asynchronous notifications to the IMF module of VCS, and
IMF (Intelligent Monitoring Framework) is the umbrella feature in VCS that can work with any 3rd-party asynchronous notification in future if required. Currently IMF uses AMF to provide the above mentioned capabilities.
- See also:
a) A screen-cast on IMF:
https://www-secure.symantec.com/connect/videos/screencast-intelligent-monitoring-framework
b) A related blog on IMF:
http://symantec.dciginc.com/2010/10/imf-sfha51sp1-proactive-monitoring.html
c) VCS User's guide (search for the keyword: IMF):
http://sfdoccentral.symantec.com/sf/5.1SP1/solaris/pdf/vcs_admin_51sp1_sol.pdf
d) A whitepaper on IMF:
https://www-secure.symantec.com/connect/articles/intelligent-monitoring-framework-imf-whitepaper
e) On a lighter note, a related quote from Einstein that elaborates on why saving 1 minute is crucial for some cases :-)
""Place your hand on a hot stove (for a minute) and it will seem to last an hour. Sit next to a beautiful lady for an hour, and it will seem to last only a minute. That is relativity." -- Einstein.
Do experiment with IMF and let us know your experiences!
Bhavin Thaker.
The Storage & Clustering Community Blog is the perfect place to share short, timely insights including product tips, news and other information relevant to the Storage & Clustering community. Any authenticated Connect member can contribute to this blog.
Comments 4 Comments • Jump to latest comment
Before IMF the agent looks the resource and as it fails VCS monitoring process declare that the resources faulted and now VCS pick the alerts directly from OS regarding the services (which it poll ) is faulted
Any comment will be appreciated. Mark as Solution if your query is resolved
__________________
Thanks in Advance
Zahid Haseeb
zahidhaseeb.wordpress.com
Hi Zahid,
In your comment:
Zahid> now VCS pick the alerts directly from OS regarding the services (which it poll ) is faulted
I would like to clarify that with IMF, polling is avoided and that when the resource is faulted the event is generated, making it truly asynchronous model of event detection.
Is there a roadmap for supported agents for IMF, like the other database agents - Informix, DB2, Sybase and also applications like WebLogic etc.
Mike
UK Symantec Consultant in VCS, GCO, SF, VVR, VxAT on Solaris, AIX, HP-ux, Linux & Windows
If this post has helped you, please vote or mark as solution
The VCS agents that support IMF are listed at: https://www-secure.symantec.com/connect/blogs/vcs-...
Would you like to reply?
Login or Register to post your comment.