What can you expect from a cloud?
Why should you prefer a cloud hosting to a traditional hosting?
The short answer: the advantages of virtualization without the inconvenient.
The detailed answer starts by explaining that a cloud is a set of physical servers, each of which is hosting many virtual machines (VMs). Each VM is an operating system (OS) on its own, so you can have many OSs running on a single server. That is the foundation of the shared hosting: many OSs, hence many applications per server. Shared hosting is cheaper than dedicated hosting, where the application runs alone on a server.
But a cloud comes with even more functionality.
Once you have provisioned your first VM (i.e. once you have installed your application on top of an OS) the cloud should enable you to replicate it and decide on the number of VMs you want to use. If provisioning the first VM can take some time you can expect the replication of the other VMs to be quick: from 1 to 30 minutes depending on the technology used for the replication. This makes the cloud an interesting solution to address a peak in workload. If you are using many VMs to run a Web application the hosting provider should also provide you with load balancing, which is needed to spread the trafic amongst the different VMs.
To get the best out of the dynamic resource allocation and out of the VM replication you need to monitor your VMs. This feature detects that one or many VMs are reaching a critical level of resource consumption. When such a situation is detected, additional resources or additional VMs can be automatically engaged.
When the server hosting the VM is saturated and you still need some additional power, the cloud should allow you to migrate the VM to a more powerful server or to a server loaded with less VMs.
Once VM migration is available, the next step is to put in place a failover mechanism: if a server crashes, the VMs running on it are automatically migrated ans restarted on another server. This operation can be accomplished in a few minutes. The failover mechanism requires that the VMs are not installed on the local disks of the server - those are not accessible anymore when the server crashes - but on a Networked Attached Storage (NAS). The NAS plays a central role in a cloud because all the precious VMs are stored on it.
For the most demanding users, it is also possible to implement High Availability on a cloud, in a way that is similar to what already exists for servers: each active VM is backed up by a passive one that is pre-configured and booted. At any time, if an active VM crashes, the passive one immediately replaces it. The only operation required is to configure the network interface of the passive VM with the IP address of the VM that crashed. With such a mechanism there is almost no service interruption - we are talking about seconds here - and the reduced cost of VMs makes HA really attractive.
Talking about HA, and considering the importance of the NAS in a cloud, it is necessary for the NAS to have its own HA mechanism in place.
Don't hesitate to talk about cloud and cloud computing with your hosting provider, and ask which functionality is currently available in its data center. The most dynamic of them certainly have many answers to provide since some of them are already deploying clouds based on Sun technologies (such as servers, Solaris zones for virtualization, and the storage 7000 for the NAS). That is the case of EveryCity in the United Kingdom and Planet-Work in France who offer scalable hosting that is tailored to your application not only in terms of computing power, but also in terms of storage.
Posted by: Thierry Manfe | Category: UK
| Comments [0]
Subscribe to the feed
