This is Part 2 of The Year of Cloud Experimentation. Please click here to read Part 1.
How is the experimentation starting?
The first steps involve application inventorying and application topology. What business applications are in the inventory? What can be migrated? How are the applications interrelated and what is their topology?
The initial cloud evaluation is going to be similar to the P2V consolidation analyses that have been occurring over the past few years. An added dimension to the Cloud assessment, besides having to understand workload profiles, is identifying proximity of data to the compute aspects of the applications. Moving large amounts of data between a Cloud and a private network is not yet feasible.
An additional element to testing the Cloud, unlike in-house virtualization, is vendor risk assessment. If a vendor does indeed collapse, how quickly can workloads be migrated to a different vendor? Are there any technological ‘gotchas’ like unsupported platforms? This will be a very important hurdle for Cloud to overcome.
Furthermore, what types of Cloud services need to be evaluated? Cloud servers, Cloud storage, dedicated hosted platforms? Looking for a DR/HA environment to duplicate in-house infrastructure? Or perhaps looking to leverage fractional compute during peak application load times? How is network infrastructure integrated? Are VPN services available? How does storage fit into the deployment plans?
Ultimately, cost modeling will be needed to determine the true cost saving. A number of factors play into this equation: compute intensity and workload profile, network demand, storage throughput, and since services will now be remote, service latency plays a larger role in application delivery.
When evaluating how to deploy applications into the Cloud, there are a number of operations issues to consider. This includes packaging applications as images for quick deployment and scaling, as well as understanding the potential patch management needs. A considerable amount of experimentation with lifecycle management tooling will be required, especially if applications are multi-tiered or distributed. Experimentation with systems management tooling will be essential in order to monitor and manage physical, virtual, and Cloud (PVC) infrastructure from a single-pane-of-glass. Will tooling integrate with automation solutions to assist in dynamic allocation of resources, or possibly avoid incidents through proactive changes in infrastructure?
Evolution in automation will occur around “bundling” workflows. Currently, there are very few sophisticated automation workflows related to Cloud technologies. For example, imagine if VMware’s Orchestrator enabled Amazon EC2 or Rackspace drag-and-drop workflows around provisioning. If this were the case, dynamic changes in demand would be possible in just a few clicks.
An area of Cloud that we don’t see yet, but which will become more relevant as Cloud services mature, is the concept of cost brokering. Ultimately, since the workloads in a private environment are known, which Cloud vendors can be used most cost-effectively for fractional compute bursts? Quite conceivably, depending on location and time of day, rates will differ and you’ll be able to take advantage of the discounts and drive concrete cost-savings.
Internally developed applications won’t be the first applications in the Cloud environment. However, it is important to evaluate which application development platforms are available in the Cloud to be future-ready. This would include environments like Engine Yard, Azure, Google AppEngine, or SpringSource.
Here is a list of the first five things that an IT manager should consider when evaluating the Cloud:
- Where on the IT Spectrum do we fit?
- What business applications do we have? What are their topologies?
- What is the profile of the application workloads (can we take advantage of fractional compute)?
- Are the applications data or network heavy? Are they highly interdependent?
- What systems management tooling passes the P-V-C (physical, virtual, and Cloud) test. Is all this infrastructure inside and outside our walls manageable through a single-pane-of-glass dashboard?
Overall, Cloud will change IT and business in a way similar to the Internet. Make no mistake, we are on the edge of a big and positive change. While there are many hurdles to overcome before the Cloud becomes a mainstream component of IT, these issues will be solved over the next few years.
Remember when the internet first started getting major traction? Were you one of the pioneers in your company that saw its potential? Don’t forget that one of the biggest software companies in the world missed the internet boat and fell behind in catching the ‘internet wave.’ Well, the horn has sounded. Cloud is the next big thing. What are you going to do with it?