With the realisation that keeping your cloud options open and having a multi-cloud (see sidebar) perspective for your applications and companies like Turbonomic Inc and Rovius Cloud providing single platforms from which to automatically manage these environments, based on your criteria, it is tempting to believe that by purchasing these tools the work is done. IT IS NOT! Whilst some of the following are true for single cloud environments they are compounded in multi-cloud
Dynamic workload placement across multiple clouds, whether they are public or private requires rules, or policies, that allow the correct decisions to be taken automatically. Since there are a huge amount of possible variables in the decisions, the creation of these policies is not a trivial task. It is also not a task that is ‘one and done’ it will need constant tweaking, and sometimes significant alteration depending on the strategy of your business.
Some of the decisions that are made by the policy engine may have been traditionally made through the creation of intelligent cloud architectures. Is this still going to be the case in your organisation? Or do the architectural rules used in cloud application design need to be relaxed in favour of autonomous operation, or vice versa?
What differences does policy based automation mean for your supplier service level management? If decisions are being taken daily, perhaps even hourly or more frequently do you trust the system from the get-go, and remove some of the metrics from your SLM calculations, what impact does this have on the SLAs you have with your consumers. As bias moves between cloud providers how do you adjust for this moving target?
And whilst the technical elements are indeed managed in large part by the multicloud management tools, a significant omission appears to be that of end-user monitoring of the application. All individual components may be working well together, but the user can still have a poor experience. How are these external factors included in any policy driven engine.
Finally, procurement. They have signed the cheque for a cloud provider, and since finance wants to be predictable and do not want varying charges throughout the year they have effectively authorised an even distribution of spend. However, the most efficient use of technology may not equate with the most efficient use of funds. Can systems trigger (and assume) approval flows outside of the management system, or must we tolerate suboptimal operation cost in favour of optimising budgeted spend? i.e once we use up all allocated budget for the best cloud provider (for a certain element) we must switch to a second choice since they have allocated budget remaining.
In the end, you’ll be OK if you keep your eyes open. Just be aware that there are more than the technological capabilities at play