Graph Similarity based Cloud Migration Service Composition Pattern Discovery

Introduction

The booming cloud computing inspires the business application migration to the Cloud. The public cloud workloads are expected to increase at CAGR (Compound Annual Growth Rate) of 50% in the next three years, according to Morgan Stanley in (Holt et al., 2011). The regional distribution of cloud workloads are expected to grow at 24% ~ 45% CAGR from 2012 to 2017, according to Cisco (Cisco, 2013). By 2016, global cloud traffic will account for nearly two-thirds of total data center traffic, according to IDC (Mahowald and Sullivan, 2012). In response to the huge migration demands, many migration service vendors have emerged in recent years, offering services based on a diverse set of migration techniques. In general, these techniques can be categorized into three major types: image-based migration, application-centric migration, and migration to virtualized containers (Wang et al., 2013). Most of vendors expose their migration ability as web services.

However, migrating existing on-premises enterprise applications to cloud is a costly, labor-intensive, and error-prone activity due to the complexity of the applications, the constraints of the clouds, and the limitations of existing migration techniques provided by migration service vendors (Frey and Hasselbring, 2010). To address these challenges, a recent paper presented an approach to find out the most cost effective solution by composing multiple migration services from different vendors together to complete one migration task (Wang et al., 2013). We call it the sandbox approach because it only provides a one-off solution by calculating without verified precedent. It adopts an exhaustive searching based algorithm with pruning to improve the efficiency. In the sandbox approach, the metric for migration service selection and composition is the total cost. In fact, the cost is not the only metric for solution selection. The reliability, privacy and other nonfunctional constraints should be considered as well. Besides, there are other nontechnical factors that will impact the selection. For example, a customer may prefer services from specific vendors. More critically, a case by case solution discovery approach has not explicitly process logic for tracing, benchmarking, debugging and optimization.

Fortunately, pattern has been proven as an appealing approach to accelerate the service composition and ensure the quality. It can also be applied to the cloud migration service composition to solve the above mentioned challenges. Pattern is based on the fact that an idea has been proven useful in one practical context and will probably be useful in others (Gamma et al., 1995). Even though, every system has its own set of prerequisites, hidden costs, one-off requirements and special case exceptions, the best practices could tell us how to cope with these issues. More specifically, service patterns are defined over services and present the typical ways of composing services to achieve certain goals (Fu et al., 2009). The pattern in this paper refers service composition pattern dedicates to the service composition for the cloud migration. Service composition patterns facilitate the web service composition and accelerate the response to the market. This is exactly why the patterns are appealing as a medium to convey solutions (Rotem-Gal-Oz, 2012).