There is a huge debate in the market about the obvious – that the battle between private dedicated deployments ‘vs’ the public cloud will only lead to a hybrid cloud. But what is hybrid cloud? Well a simple straight answer is that its a mix of public and private/dedicated cloud.

It is clear that SMB markets and Enterprises will eventually (infact its already happening) are looking for ways to deploy their test/dev environments on commodity cheap hardware while production has been deployed on high performance hardware where support/performance/DR is guaranteed(atleast for the most part). And again it also depends on splitting the work loads – would you like to scale your web environments across multiple servers – which you can deploy in an instant using an API over the cloud while high performance database and application servers run on your dedicated gear. But does this seemingly simple setup really qualify as “true hybrid cloud”?

In my view – true hybrid cloud should be defined as –

“The ability to deploy, access, control and migrate between public and dedicated infrastructure using a single API”.

The real idea is to use a single API to control both public and dedicated footprints. Why a single API? Think about it – can you gauge how many development cycles will saved if you had one API to deal with. With that one API you could spin up virtual machines on both public and private/dedicated clouds – you could change networks on the fly and – this is the real sweet – “migrate” a virtual machine from public to private cloud without breaking a sweat.

If you are rolling your eyes after reading the word “migrate” – well its already being worked on! If you frequent my blog you surely must have read about NetApp Shift. NetApp Shift uses flex clone technology to literally migrate a virtual machine from one hypervisor to another. It currently supports VMware, Hyper-V and Citrix Xen Server but Vaughan Stewart of the NetApp world has said that more hypervisors will be added to the portfolio. I am certain that OpenStack will be added in the future and with that – truly this opens the doors to the above “migration” using storage array based technology.

And I am not done yet! VMware has already released the second version of multi-hypervisor manager. Currently this manages VMware and Hyper-V but this is surely VMware’s gateway to add Openstack as well. If you are still wondering about a real use case of migrating a virtual machine from one hypervisor to another – and in essence from public to private/dedicated foot print – think of a pushing a development project into production without having to rebuild it from scratch. Develop in public cloud – cheap and fast – and push it to private dedicated space when ready for production. Another one is your seamless adoption of public cloud over private footprint – do you want to rebuild all your virtual machines or simply migrate them? How about spinning up remote desktops (VDI – virtual desktop interface) with the ability to migrate them to public cloud if you find the user is logged on for prolonged periods of time or vice-versa? These are however only a few use cases, there are many more and you are only limited by your imagination.

The strategic side of me says that VMware may release an API for its multi-hypervisor manager which will allow you to manage multiple hypervisors across a set of vendors with that one API. VMware will secure vendor lock in with that. I mean really at that point – unless you develop code to pass around json and xml, dumping VMware’s API might require lot of rework. From a business side of things – dumping a vendor is not a big deal. Its all about what does that one API really give you. Its all about flexibility and the ability to use that one API among multiple vendors and spread your risk across them. Think of “mutual fund” here – that makes for a good analogy!

My emphasis however is to use a more vendor nuetral API to manage your hybrid cloud infrastructure. Openstack is an obvious choice. OpenStack today has drivers that will allow you to talk to VMware virtual center. I agree they are very basic but its a start. With OpenStack – you can spread your foot prints across multiple hosting providers and this goes back to spreading your risk as mentioned above. You could have hosting provider A for your public/commodity cloud while hosting provider B can be your dedicated foot print. Heck you could host your own dedicated foot print and have a public cloud foot print else where and use “one API” to run them both.

Another good choice would be Delta Cloud – which has its roots with RedHat and the Apache foundation. Delta cloud basically abstracts  the cloud differences and provides a REST API that allows you to interact between multiple cloud providers. Delta cloud has drivers for Amazon EC2, Rackspace, OpenNebula, Terremark and VMware vCloud are a few. However I don’t see much information on if Delta cloud supports inter-op between public IAAS cloud providers and a private/dedicated footprint and by private here I mean – straight support for VMware/Hyper-v/citrix.

I am working on a demo of Openstack managing vCenter and will post it soon.

Do comment and critic as needed.


Leave a Reply

Your email address will not be published.

Post Navigation