Wind of Change: from Vendor Lock-in to the Meta Cloud

The emergence of more and more cloud offerings from a multitude of service providers calls for a meta cloud, which smoothens the jagged cloud landscape. We discuss our proposal for such a meta cloud, and explain how it solves the lock-in problems that current users of public and hybrid clouds face. The cloud computing paradigm has found widespread adoption throughout the last years. The reason for the success of cloud computing is the possibility to use services on-demand with a pay as you go pricing model, which proved to be convenient in many respects. Because of low costs and high flexibility, migrating to the cloud is indeed compelling. Despite the obvious advantages of cloud computing, many companies hesitate to “move into the cloud”, mainly because of concerns related to availability of service, data lock-in, and legal uncertainties [1]. Lock-in is particularly problematic for the following reasons. Firstly, even though availability of public clouds is generally high, eventual outages still occur [3]. If this is the case, businesses locked into such a cloud are essentially at a standstill until the cloud is back online. Secondly, public cloud providers generally do not guarantee particular service level agreements [4], i.e., businesses locked into a cloud have no guarantees that this cloud will continue to provide the required Quality of Service (QoS) tomorrow. Thirdly, the terms of service of most public cloud providers allow the provider to unilaterally change pricing of their service at any time. Hence, a business locked in a cloud has no midor long-term control over their own IT costs. At the core of all of these problems, we can identify a need for businesses to permanently monitor the cloud they are using, and to be able to rapidly “change horses”, i.e., migrate to a different cloud if monitoring discovers problems or estimations foresee issues in the future. However, migration is currently far from trivial. A plethora of cloud providers is flooding the market with a confusing body of services, such as compute services e.g., Amazon Elastic Compute Cloud (EC2) and VMware vCloud, or key-value stores, e.g., Amazon Simple Storage Service (S3). Evidently, some of these services are conceptually comparable to each other, others are vastly different, but all of them are, ultimately, technically incompatible and follow no standards but