Open Stack

OpenStack and other cloud products are more about resiliency and less 
about High Availabilty.

HA is about making sure the systems don't go down because the cost of 
going down is catastrophic.
Think Space shuttles or aircraft or Nuclear power generating stations.
With HA you try and find all single points of failure and remove them.
So you need double or triple components and subsystems.
Networks, power supplies, fencing devices.

Most folks screw up HA on things like fencing devices.
You need at least 2 different fencing technologies and need to make sure 
all your equipment fenceable.
There is nothing worse than having a single component failure bring your 
HA infrastructure to its knees.

I have had file systems completely corrupted beyond recovery by a few 
seconds of operating with a failed fencing device.

If your serious about HA you need to have a serious failure mode analysis.

Most folks don't do all this.
They get DRBD and a couple of machines or Proxmox and put them up with 
no fencing or redundant networks.
And life is good til the first corner case failure.

My rule of thumb for HA costs is  2 x reliability == 10 x cost

Resiliency is about knowing that there will be failures but designing 
the system to be tolerant of component failure.
Allowing reduced or degraded performance while the broken systems are 
removed or repaired.

So you use something like Ceph. A low cost distributed data store that 
is designed to allow for multiple component failures but the overall 
file store will continue to operate.

With OpenStack the idea is that you have extra hardware and push the 
load around to whatever hardware can accommodate it.
If a server fails its no big deal because you have N more where you can 
share the load of the migrated processes.
It is easy to add another server to the pool of servers so that you can 
move the load around.

My personal feeling is that in most cases of utility computing 
Resiliency gives better reliability than HA.

HA is good for locations where the application can be managed and is not 
extremely dynamic.
I use it for core routers because its a small constrained environment 
where its easy to control the power and networks.

HA has its place but I would argue that its not as a general purpose 
cloud platform.

There is my $.02

On 05/13/2014 11:57 AM, Frans Thamura wrote:
> Hi all
>
> we have a little homework for one of infra teacher here, which he
> master proxmos, regarding openstack as default education vs current
> proxmox.
>
> i said about controller and HA,
>
> but Proxmos have HA also.
>
> this is his HA
>
> Proxmox VE High Availability Cluster
>
> Proxmox VE HA Cluster enables the definition of high available virtual
> servers. If a virtual machine or container (VM or CT) is configured as
> HA and the physical host fails, the VM is automatically restarted on
> one of the remaining Proxmox VE Cluster nodes.
>
> The Proxmox VE HA Cluster is based on proven Linux HA technologies,
> providing stable and reliable HA service.
>
>
>
> anyone can help to confince him?
>
>
> F
>
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack


-- 
Alvin Starr                   ||   voice: (905)513-7688
Netvel Inc.                   ||   Cell:  (416)806-0133
alvin at netvel.net              ||