[ubuntu-cloud] UEC Single Point of failure

[Zeeshan Ali Shah]

I assume Amazon has same kind of notion "CLC" , how they managed to get 
the load balance so effictevely ?

Zeeshan

On 11/04/2010 09:23 AM, Torsten Spindler wrote:
> On Thu, 2010-11-04 at 09:13 +0100, Björn Böttcher wrote:
> ...
>
>    
>> Still open question:
>> But do the multiple CC/SC machines synchronize automatically?
>>      
> Nope, the cluster controllers (CC) manage different clusters and do not
> synchronize any state between them. Same for the storage controllers
> (SC) that are attached to different clusters.
>
> Instead your payload on UEC (e.g. a webapp) needs to be tailored for
> fault tolerance, when a cluster fails. For example, a load balancer
> outside of UEC may direct requests to either cluster in your UEC. If one
> fails, the load balancer learns about it (maybe through a heartbeat?)
> and directs any requests to the remaining cluster(s).
>
> On your original question, the cloud controller (CLC) is indeed a single
> point of failure. Putting it on hardware that provides some resilience
> towards failures is best practice there.
>
> Torsten
>
>
>
>    


-- 
Regards

Zeeshan Ali Shah
System Administrator
PDC-Center for High Performance Computing
KTH-Royal Institute of Technology , Sweden
+46 8 790 9115