“A Bird’s-Eye Look at Cloud Computing offerings” published on November, 26th, generated a lot of interest from the Cloud providers. Both mentioned providers contacted me, providing some corrections to the blog post and requesting some additional feedback. Even a new Cloud start-up has requested some advice and offered me to be part of the beta program.
[email protected]: #Cloud computing is still being invented; any other view of it generates a palm to my forehead chain reaction.”
On the same line of thinking I came up some features, not too far from today’s technology, that we would like to see available for the next Cloud release.
During my tests in A Bird’s-Eye Look at Cloud Computing offerings I was able to select a template to be deployed and ‘god knows’ where those virtual disks were deployed in the disk array. Just like we do storage layering internally in organisations it would be nice to have the ability to choose what type of RAID and performance virtual disks should have. This feature would make databases and log intensive applications much more favorable to be migrated to the Cloud.
The providers also should mention characteristics and performance of the arrays.
In order to maximise CPU core utilisation Cloud providers do not specify the amount of CPU cycles or GHz assigned to each VM. That is how they intend to make money, however organisations will need a guarantee that their workloads will receive enough CPU cycles when needed; and this is most likely to happen at the same time that all other organisations need CPU cycles.
It is crystal clear to me that without this warranty the most critical workloads will remain inside the organisations for a long time.
Most Cloud operating systems provide capabilities such as CPU affinity and resource pools, however how the providers are going to charge for the service is the million dollar question. Would be fair to get more CPU cycles only if you pay more?
The truth is that Cloud providers that can guarantee CPU cycles will have to invest more therefore their service will cost more.
VM Portability for compatible Clouds
This is not too far from happening but would be handy to have the ability to transfer VMs from internal infrastructure to the Cloud and vice-versa. At the end of the day there are 3 or 4 different Cloud OS’s, so if you run VMware in your internal environment you should be able to Storage vMotion a VM to the Cloud and from the Cloud.
During vForum’09 there was an interesting session with Rodney Haywood (@rodos) and Optus demonstrating VMware vCenter connected to the internal virtual infrastructure and to the Cloud at the same time.
Once this is accomplished than we can then start dreaming about Cloud -to-Cloud portability and Open Standards for Cloud interoperability.
Some other ‘wishes’ but a bit further down the track:
- Application Driven VM (VMware’s dream and the reason for Spring Source acquisition – the application will drive the infrastructure deciding when to upgrade memory, CPU etc.)
- Live Cloud migration (Non disruptive VM migration between Clouds)
- vAPP Environment Taxonomy (vAPP will have the ability to specify the environment that a group of VMs will run at, such as CPU, RAM, Bandwidth etc.)
What else would you like to include in this wish list?