After the mad rush to virtualise, what then?
Tools for building the hybrid cloud
Cloud Those of us who have set up local clouds had to go through a transition period; some systems dedicated, some systems virtualised. For many, running a mix of dedicated and virtual systems proved advantageous.
The mad rush to virtualise everything slowed. Low hanging fruit was virtualised yielding beneficial returns. Difficult systems or those with unique requirements were left on dedicated hardware. It was a simple case of the drawbacks outweighing the benefits.
Embracing the hosted cloud has a similar transition point; some systems local, some systems hosted. Occupying this in-between state is the hybrid cloud. Here, the focus is integration between local and hosted services.
Information from one service must reach the next, even if that service is not located on the same network. In some cases this extends to the ability to enable, track, and cope with dynamic workload migration between local and hosted providers.
Services bridging the gap between clouds must know where the various elements are living, and what conditions exist between them. Attention must be paid to bandwidth constraints, latency and service outages.
Unpredictable, that's what you are
Avoiding the pitfalls of hybrid cloud computing is more than just good coding. A new generation of applications is emerging which make direct use of cloud APIs. These applications don’t rely on virtual machines for environmental awareness or dynamic workload migration. Awareness and tolerance of unpredictable bandwidth conditions is critical if the hybrid cloud applications are to succeed.
Applications coded for their platforms can be run in compatible local clouds or migrated out to hosted providers as needed. Hybrid cloud applications typically run with an instance on each cloud, dynamically moving loads for best fit. Applications that are not as cloud-aware must rely on virtual machines. Here, management tools are absolutely critical to realising the promise of the hybrid cloud.
The theory goes that workloads can be migrated from the local cloud onto far more powerful hosted clouds automatically. For predictable workloads, migrations can be scheduled. When dealing with the unexpected, workloads migration is initiated when activity rises above predefined limits.
The tools available are as varied as the cloud strategies of the vendors underpinning them. VMware’s vCloud Director is an excellent example of a virtual-machine centric approach. Moving virtual machines between a local cloud and a hosted cloud proved simple. Amazon’s weighs in with VMware Import, including a plug-in for vCenter. Intel is spinning a new hybrid cloud – currently in closed beta – promising similar functionality.
The environmental awareness of your applications will be a big factor here. Consider the remote computing application Teamviewer. You establish a connection not directly with the target PC, but instead through an HTTP tunnel bouncing off the Teamviewer public cloud. Compare this to an application so old that it can only access its database or storage if it can resolve the name using WINS.
If your application doesn’t make proper use of DNS, publicly routable address space and encryption, all is not lost. There are hybrid-aware virtual private networking tools available to solve this. Management tools – through add-ins – are capable of reconfiguring the guest VM’s networking. If you have the budget, your ISP may offer easy options such as MPLS to connect your local network with hosted services.
Application interoperability can also become a casualty of the hybrid cloud via other routes. Managed cloud services often include behind-the-scenes application maintenance by the hosted provider. The application of a patch or an upgrade to a new version can trigger issues from changes in the default format of exported files to newer versions no longer even supporting the formats or protocols you require.
The transition to a hybrid computing model begins with small steps. Your first project should be simple; migrating to a hosted Exchange provider perhaps, or incorporating hosted VMware services into your disaster recovery plan. The hybrid cloud model has its hang-ups, but once you figure them out it’s worth it. ®
Trevor Pott is a sysadmin for a small-ish company in Edmonton, Canada.