Optimization of Web Services for Cloud Deployment and Mobile Consumption

Abstract

Research performed for this thesis indicates an impedance mismatch between prevailing approaches to development of service-oriented enterprise applications and the consumption capabilities of mobile devices. The rich semantics and strong validation mechanisms inherent in SOAP-based web services, common to large-scale enterprise development, introduce inefficiencies of network bandwidth consumption and serialization/de-serialization processing requirements. These inefficiencies may be financially burdensome when systems are migrated to a cloud-based hosting environment and both costly and non-performant when accessed from network and processor constrained mobile devices. Yet wholesale abandonment of established enterprise practice and legacy systems for the adoption of unfamiliar architectural styles is rarely practical. This thesis proposes a series of incremental changes to enterprise web services architecture that, individually, provide measurable efficiency benefits both when served from the cloud and when consumed from mobile devices. The objective of this research is to quantify the benefits and illustrate trade-offs for each. Within a cloud deployment, selective application of HTTP compression is shown to yield performance improvements in excess of 40% with data transfer reductions of up to 85%. Analysis identifies the characteristics of services that suffer degraded performance under compression, and illustrates how similar performance and data reduction benefits may be achieved through service augmentation with alternative message and request formats. Thesis focus then turns to options for improving efficiency in the consumption of these services from native applications on prevailing mobile device platforms. Development and measurements performed for this thesis identify approaches for faster and more efficient processing of existing services on mobile devices and relates these to the developer effort required. Further enhancements to application performance and development simplicity are demonstrated through mobile consumption of the augmented services and formats proposed for optimized cloud deployment. Research for this thesis suggests that in both cloud and mobile sides of a distributed system, performance and financial benefits may be achieved while building upon, rather than replacing, existing services code and architectural patterns.