[fedora-virt] KFT feature proposal review for F17

From: Anand Nande <anande@redhat.com>
To: Development discussions related to Fedora
	<devel@lists.fedoraproject.org>,
	fedora-virt@redhat.com
Cc: xen@lists.fedoraproject.org, virt@lists.fedoraproject.org,
	xen-devel@lists.xensource.com, xen-users@lists.xensource.com
Subject: [fedora-virt]  KFT feature proposal review for F17
Date: Thu, 15 Sep 2011 19:20:32 +0530	[thread overview]
Message-ID: <4E7202A8.5060600@redhat.com> (raw)
In-Reply-To: <20110914112552.GP12984@reaktio.net>

[-- Attachment #1.1: Type: text/plain, Size: 2286 bytes --]

KFT (KVM Faul Tolernace):
========================

How it works:
-------------
* Provide Continuous availability to Mission Critical Applications with 
a 'Always Available' environment preventing any downtime or Data-loss in 
the event of server-failures.
* RFT when enabled for a Virtual Machine creates a secondary copy of it 
on another physical server running KVM.
* The two copies are continuously synchronized by replaying the contents 
of the Memory from the primary VM to the Secondary (copy) VM. The 
Primary and the Secondary VM access the same storage device (shared 
storage) between the 2 servers.
* For example: If a user is playing a Youtube video on VM1, and the KVM 
Guest VM-1 crashes (due to some reason), The Secondary-slave copy 
becomes Active and promotes itself as Primary immediately when the 
vdsm-agent (communicating between the KVM-Host and the KVM-GuestVM-1) 
detects the server crash. This leads to creation of another secondary 
copy on the Next available physical server (obviously sharing the same 
storage with the now-active Primary-VM).
* Thus the GuestVM-2 (which was the secondary copy before the crash) 
---> is now known as VM-1 (and the new secondary copy on the newly 
available server is now known as VM-2).
* This can be achieved over qemu+ssh connections being managed by a 
single virt-manager as well.

Advantages:
-----------
* Zero Downtime: Fault Tolerance provides absolutely no downtime as that 
comapred to HA wherein the VM restarts completely on a different server.
* Immunity from Server crashes/panic/OOM situations.
* Extremely Fast Disaster Recover Mechanism.

Disadvantages:
--------------
* If a panic/crash takes place on the GuestVM being hosted on the KVM, 
This will be replicated on its Secondary-Copy, which is being hosted on 
other KVM's.
   [ This can easily be avoided by having periodic snapshots of the 
Primary-VM ]

Currently Available Technologies:
---------------------------------
* Kemari: Fault Tolerance

Xen source: 
http://www.xen.org/files/xensummitboston08/tamura_xen_summit_presentation_final.pdf
KVM source: 
http://www.linux-kvm.org/wiki/images/0/0d/0.5.kemari-kvm-forum-2010.pdf
Kemari Project: http://www.osrg.net/kemari/

Please leave your valuable feedback for this.

Sincerely,
Anand Nande

[-- Attachment #1.2: Type: text/html, Size: 3605 bytes --]

[-- Attachment #2: Type: text/plain, Size: 148 bytes --]

_______________________________________________
virt mailing list
virt@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/virt