When we started out to architect Sentinel Cloud 2 ½ years ago Brewer's CAP theorem helped a lot to get the people's mindset straight. We knew the network does fail and customers need high availability, so we drilled down what weak consistency means for our offering. And that really was driving a revolution in how to think about licensing and authorization in the cloud versus licensing for on-premise. Traditionally people think about licensing like one central control point having all data and always having full visibility. As an example take a typical enterprise concurrency license model, where a software publisher wants to know how many concurrent licenses are in use – now take this model to the cloud... boom! you are doomed to fail. The service consumers will complain about slow response times, as all geo's will need to talk to one global instance, managing all sessions to get to a strictly consistent global view. In addition this one system is the perfect example for a single point of failure – boom! again.
So we took the path to completely replace 'old technology' and built our system to scale horizontally, globally and support weak, "eventual" consistency (like Amazon calls it). We have transformed the traditional enterprise license models to eventually consistent license models, employing asynchronous communication, to propagate state to stateless service nodes performing the decomposed functions required to collect, aggregate, prepare and distribute the information. Adding to this, local intelligence with smart caching enables us to deliver snappy response times and the means to cope with load and failures – something we never could have gotten to with the traditional architecture.
But I think there is another very important take away. From an offering perspective every company moving to the cloud needs to consider, if their traditional offering does fit a cloud model – in case you have relied on strict consistency you will need to have a careful look if you really need it as it makes achieving all the design goals Tim mentioned very difficult, partially even impossible. So moving to the cloud is beyond us doing the right things in technology also our business companions need to understand the implications of walking in the clouds... Michael "MiZu" Zunke, CTO SRM, SafeNet Inc.