As distributed, global-scale, data-intensive applications are becoming more and more common, an increasing pressure is being put on the underlying distributed data services. As such services need to support massively concurrent, largely distributed accesses to huge shared datasets, the stability and scalability of their performance are critical. More specifically, the ability to sustain a stable high throughput is a very desirable property, as it strongly impacts the quality of service of the data storage system and thereby the overall application performance. Handling quality of service in a large-scale distributed system is however a very difficult task, as a very large number of factors are involved: the data access patterns, the status of a huge number of physical components, etc. In this paper we explore an approach to the management of quality of service in distributed storage systems based on global behavior modeling combined with client-side quality of service feedback. Our objective is to automate the process of identifying dangerous behavior patterns in storage services. To demonstrate our approach, we apply GloBeM, a global behavior modeling technique based on monitoring data analysis and machine learning, to improve the quality of service in BlobSeer, a distributed storage for large-scale data-intensive applications specifically designed to sustain high throughput under heavy access concurrency. We evaluate this improvement through extensive evaluations on the Grid'5000 testbed using hard experimental conditions: highly-concurrent data access patterns, for long periods of service uptime, while supporting failures of the physical storage components. Our results show substantial progress in sustaining a higher and more stable data access throughput.